diff options
Diffstat (limited to '')
| -rw-r--r-- | libjpegtwrp/jdcoefct.c | 1038 |
1 files changed, 1038 insertions, 0 deletions
diff --git a/libjpegtwrp/jdcoefct.c b/libjpegtwrp/jdcoefct.c new file mode 100644 index 000000000..84ce56fb7 --- /dev/null +++ b/libjpegtwrp/jdcoefct.c @@ -0,0 +1,1038 @@ +/* + * jdcoefct.c + * + * Copyright (C) 1994-1997, Thomas G. Lane. + * This file is part of the Independent JPEG Group's software. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains the coefficient buffer controller for decompression. + * This controller is the top level of the JPEG decompressor proper. + * The coefficient buffer lies between entropy decoding and inverse-DCT steps. + * + * In buffered-image mode, this controller is the interface between + * input-oriented processing and output-oriented processing. + * Also, the input side (only) is used when reading a file for transcoding. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" + +/* Block smoothing is only applicable for progressive JPEG, so: */ +#ifndef D_PROGRESSIVE_SUPPORTED +#undef BLOCK_SMOOTHING_SUPPORTED +#endif + +/* Private buffer controller object */ + +typedef struct { + struct jpeg_d_coef_controller pub; /* public fields */ + + /* These variables keep track of the current location of the input side. */ + /* cinfo->input_iMCU_row is also used for this. */ + JDIMENSION MCU_ctr; /* counts MCUs processed in current row */ + int MCU_vert_offset; /* counts MCU rows within iMCU row */ + int MCU_rows_per_iMCU_row; /* number of such rows needed */ + + /* The output side's location is represented by cinfo->output_iMCU_row. */ + + /* In single-pass modes, it's sufficient to buffer just one MCU. + * We allocate a workspace of D_MAX_BLOCKS_IN_MCU coefficient blocks, + * and let the entropy decoder write into that workspace each time. + * (On 80x86, the workspace is FAR even though it's not really very big; + * this is to keep the module interfaces unchanged when a large coefficient + * buffer is necessary.) + * In multi-pass modes, this array points to the current MCU's blocks + * within the virtual arrays; it is used only by the input side. + */ + JBLOCKROW MCU_buffer[D_MAX_BLOCKS_IN_MCU]; + +#ifdef D_MULTISCAN_FILES_SUPPORTED + /* In multi-pass modes, we need a virtual block array for each component. */ + jvirt_barray_ptr whole_image[MAX_COMPONENTS]; +#endif + +#ifdef BLOCK_SMOOTHING_SUPPORTED + /* When doing block smoothing, we latch coefficient Al values here */ + int * coef_bits_latch; +#define SAVED_COEFS 6 /* we save coef_bits[0..5] */ +#endif +} my_coef_controller; + +typedef my_coef_controller * my_coef_ptr; + +/* Forward declarations */ +METHODDEF(int) decompress_onepass + JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf)); +#ifdef D_MULTISCAN_FILES_SUPPORTED +METHODDEF(int) decompress_data + JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf)); +#endif +#ifdef BLOCK_SMOOTHING_SUPPORTED +LOCAL(boolean) smoothing_ok JPP((j_decompress_ptr cinfo)); +METHODDEF(int) decompress_smooth_data + JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf)); +#endif + + +LOCAL(void) +start_iMCU_row (j_decompress_ptr cinfo) +/* Reset within-iMCU-row counters for a new row (input side) */ +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + + /* In an interleaved scan, an MCU row is the same as an iMCU row. + * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows. + * But at the bottom of the image, process only what's left. + */ + if (cinfo->comps_in_scan > 1) { + coef->MCU_rows_per_iMCU_row = 1; + } else { + if (cinfo->input_iMCU_row < (cinfo->total_iMCU_rows-1)) + coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor; + else + coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height; + } + + coef->MCU_ctr = 0; + coef->MCU_vert_offset = 0; +} + + +/* + * Initialize for an input processing pass. + */ + +METHODDEF(void) +start_input_pass (j_decompress_ptr cinfo) +{ + cinfo->input_iMCU_row = 0; + start_iMCU_row(cinfo); +} + + +/* + * Initialize for an output processing pass. + */ + +METHODDEF(void) +start_output_pass (j_decompress_ptr cinfo) +{ +#ifdef BLOCK_SMOOTHING_SUPPORTED + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + + /* If multipass, check to see whether to use block smoothing on this pass */ + if (coef->pub.coef_arrays != NULL) { + if (cinfo->do_block_smoothing && smoothing_ok(cinfo)) + coef->pub.decompress_data = decompress_smooth_data; + else + coef->pub.decompress_data = decompress_data; + } +#endif + cinfo->output_iMCU_row = 0; +} + + +/* + * Decompress and return some data in the single-pass case. + * Always attempts to emit one fully interleaved MCU row ("iMCU" row). + * Input and output must run in lockstep since we have only a one-MCU buffer. + * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED. + * + * NB: output_buf contains a plane for each component in image, + * which we index according to the component's SOF position. + */ + +METHODDEF(int) +decompress_onepass (j_decompress_ptr cinfo, JSAMPIMAGE output_buf) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + JDIMENSION MCU_col_num; /* index of current MCU within row */ + JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1; + JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1; + int blkn, ci, xindex, yindex, yoffset, useful_width; + JSAMPARRAY output_ptr; + JDIMENSION start_col, output_col; + jpeg_component_info *compptr; + inverse_DCT_method_ptr inverse_DCT; + +#ifdef ANDROID_TILE_BASED_DECODE + if (cinfo->tile_decode) { + last_MCU_col = + (cinfo->coef->MCU_column_right_boundary - + cinfo->coef->MCU_column_left_boundary) - 1; + } +#endif + + /* Loop to process as much as one whole iMCU row */ + for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row; + yoffset++) { + for (MCU_col_num = coef->MCU_ctr; MCU_col_num <= last_MCU_col; + MCU_col_num++) { + /* Try to fetch an MCU. Entropy decoder expects buffer to be zeroed. */ + if (MCU_col_num < (unsigned int)coef->pub.MCU_columns_to_skip) { + (*cinfo->entropy->decode_mcu_discard_coef) (cinfo); + continue; + } else { + jzero_far((void FAR *) coef->MCU_buffer[0], + (size_t) (cinfo->blocks_in_MCU * SIZEOF(JBLOCK))); + if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) { + /* Suspension forced; update state counters and exit */ + coef->MCU_vert_offset = yoffset; + coef->MCU_ctr = MCU_col_num; + return JPEG_SUSPENDED; + } + } + /* Determine where data should go in output_buf and do the IDCT thing. + * We skip dummy blocks at the right and bottom edges (but blkn gets + * incremented past them!). Note the inner loop relies on having + * allocated the MCU_buffer[] blocks sequentially. + */ + blkn = 0; /* index of current DCT block within MCU */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + /* Don't bother to IDCT an uninteresting component. */ + if (! compptr->component_needed) { + blkn += compptr->MCU_blocks; + continue; + } + inverse_DCT = cinfo->idct->inverse_DCT[compptr->component_index]; + useful_width = (MCU_col_num < last_MCU_col) ? compptr->MCU_width + : compptr->last_col_width; + output_ptr = output_buf[compptr->component_index] + + yoffset * compptr->DCT_scaled_size; + start_col = MCU_col_num * compptr->MCU_sample_width; + for (yindex = 0; yindex < compptr->MCU_height; yindex++) { + if (cinfo->input_iMCU_row < last_iMCU_row || + yoffset+yindex < compptr->last_row_height) { + output_col = start_col; + for (xindex = 0; xindex < useful_width; xindex++) { + (*inverse_DCT) (cinfo, compptr, + (JCOEFPTR) coef->MCU_buffer[blkn+xindex], + output_ptr, output_col); + output_col += compptr->DCT_scaled_size; + } + } + blkn += compptr->MCU_width; + output_ptr += compptr->DCT_scaled_size; + } + } + } + /* Completed an MCU row, but perhaps not an iMCU row */ + coef->MCU_ctr = 0; + } + /* Completed the iMCU row, advance counters for next one */ + cinfo->output_iMCU_row++; + if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) { + start_iMCU_row(cinfo); + return JPEG_ROW_COMPLETED; + } + /* Completed the scan */ + (*cinfo->inputctl->finish_input_pass) (cinfo); + return JPEG_SCAN_COMPLETED; +} + + +/* + * Dummy consume-input routine for single-pass operation. + */ + +METHODDEF(int) +dummy_consume_data (j_decompress_ptr cinfo) +{ + return JPEG_SUSPENDED; /* Always indicate nothing was done */ +} + +#ifdef D_MULTISCAN_FILES_SUPPORTED +/* + * Consume input data and store it in the full-image coefficient buffer. + * We read as much as one fully interleaved MCU row ("iMCU" row) per call, + * ie, v_samp_factor block rows for each component in the scan. + * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED. + */ + +METHODDEF(int) +consume_data (j_decompress_ptr cinfo) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + JDIMENSION MCU_col_num; /* index of current MCU within row */ + int blkn, ci, xindex, yindex, yoffset; + JDIMENSION start_col; + JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN]; + JBLOCKROW buffer_ptr; + jpeg_component_info *compptr; + + /* Align the virtual buffers for the components used in this scan. */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + buffer[ci] = (*cinfo->mem->access_virt_barray) + ((j_common_ptr) cinfo, coef->whole_image[compptr->component_index], + cinfo->tile_decode ? 0 : cinfo->input_iMCU_row * compptr->v_samp_factor, + (JDIMENSION) compptr->v_samp_factor, TRUE); + /* Note: entropy decoder expects buffer to be zeroed, + * but this is handled automatically by the memory manager + * because we requested a pre-zeroed array. + */ + } + unsigned int MCUs_per_row = cinfo->MCUs_per_row; +#ifdef ANDROID_TILE_BASED_DECODE + if (cinfo->tile_decode) { + int iMCU_width_To_MCU_width; + if (cinfo->comps_in_scan > 1) { + // Interleaved + iMCU_width_To_MCU_width = 1; + } else { + // Non-intervleaved + iMCU_width_To_MCU_width = cinfo->cur_comp_info[0]->h_samp_factor; + } + MCUs_per_row = jmin(MCUs_per_row, + (cinfo->coef->column_right_boundary - cinfo->coef->column_left_boundary) + * cinfo->entropy->index->MCU_sample_size * iMCU_width_To_MCU_width); + } +#endif + + /* Loop to process one whole iMCU row */ + for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row; + yoffset++) { + // configure huffman decoder +#ifdef ANDROID_TILE_BASED_DECODE + if (cinfo->tile_decode) { + huffman_scan_header scan_header = + cinfo->entropy->index->scan[cinfo->input_scan_number]; + int col_offset = cinfo->coef->column_left_boundary; + (*cinfo->entropy->configure_huffman_decoder) (cinfo, + scan_header.offset[cinfo->input_iMCU_row] + [col_offset + yoffset * scan_header.MCUs_per_row]); + } +#endif + + // zero all blocks + for (MCU_col_num = coef->MCU_ctr; MCU_col_num < MCUs_per_row; + MCU_col_num++) { + /* Construct list of pointers to DCT blocks belonging to this MCU */ + blkn = 0; /* index of current DCT block within MCU */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + start_col = MCU_col_num * compptr->MCU_width; + for (yindex = 0; yindex < compptr->MCU_height; yindex++) { + buffer_ptr = buffer[ci][yindex+yoffset] + start_col; + for (xindex = 0; xindex < compptr->MCU_width; xindex++) { + coef->MCU_buffer[blkn++] = buffer_ptr++; +#ifdef ANDROID_TILE_BASED_DECODE + if (cinfo->tile_decode && cinfo->input_scan_number == 0) { + // need to do pre-zero ourselves. + jzero_far((void FAR *) coef->MCU_buffer[blkn-1], + (size_t) (SIZEOF(JBLOCK))); + } +#endif + } + } + } + + + /* Try to fetch the MCU. */ + if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) { + /* Suspension forced; update state counters and exit */ + coef->MCU_vert_offset = yoffset; + coef->MCU_ctr = MCU_col_num; + return JPEG_SUSPENDED; + } + } + /* Completed an MCU row, but perhaps not an iMCU row */ + coef->MCU_ctr = 0; + } + /* Completed the iMCU row, advance counters for next one */ + if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) { + start_iMCU_row(cinfo); + return JPEG_ROW_COMPLETED; + } + /* Completed the scan */ + (*cinfo->inputctl->finish_input_pass) (cinfo); + return JPEG_SCAN_COMPLETED; +} + +/* + * Consume input data and store it in the coefficient buffer. + * Read one fully interleaved MCU row ("iMCU" row) per call. + */ + +METHODDEF(int) +consume_data_multi_scan (j_decompress_ptr cinfo) +{ + huffman_index *index = cinfo->entropy->index; + int i, retcode, ci; + int mcu = cinfo->input_iMCU_row; + jinit_phuff_decoder(cinfo); + for (i = 0; i < index->scan_count; i++) { + (*cinfo->inputctl->finish_input_pass) (cinfo); + jset_input_stream_position(cinfo, index->scan[i].bitstream_offset); + cinfo->output_iMCU_row = mcu; + cinfo->unread_marker = 0; + // Consume SOS and DHT headers + retcode = (*cinfo->inputctl->consume_markers) (cinfo, index, i); + cinfo->input_iMCU_row = mcu; + cinfo->input_scan_number = i; + cinfo->entropy->index = index; + // Consume scan block data + consume_data(cinfo); + } + cinfo->input_iMCU_row = mcu + 1; + cinfo->input_scan_number = 0; + cinfo->output_scan_number = 0; + return JPEG_ROW_COMPLETED; +} + +/* + * Same as consume_data, expect for saving the Huffman decode information + * - bitstream offset and DC coefficient to index. + */ + +METHODDEF(int) +consume_data_build_huffman_index_baseline (j_decompress_ptr cinfo, + huffman_index *index, int current_scan) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + JDIMENSION MCU_col_num; /* index of current MCU within row */ + int ci, xindex, yindex, yoffset; + JDIMENSION start_col; + JBLOCKROW buffer_ptr; + + huffman_scan_header *scan_header = index->scan + current_scan; + scan_header->MCU_rows_per_iMCU_row = coef->MCU_rows_per_iMCU_row; + + size_t allocate_size = coef->MCU_rows_per_iMCU_row + * jdiv_round_up(cinfo->MCUs_per_row, index->MCU_sample_size) + * sizeof(huffman_offset_data); + scan_header->offset[cinfo->input_iMCU_row] = + (huffman_offset_data*)malloc(allocate_size); + index->mem_used += allocate_size; + + huffman_offset_data *offset_data = scan_header->offset[cinfo->input_iMCU_row]; + + /* Loop to process one whole iMCU row */ + for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row; + yoffset++) { + for (MCU_col_num = coef->MCU_ctr; MCU_col_num < cinfo->MCUs_per_row; + MCU_col_num++) { + // Record huffman bit offset + if (MCU_col_num % index->MCU_sample_size == 0) { + (*cinfo->entropy->get_huffman_decoder_configuration) + (cinfo, offset_data); + ++offset_data; + } + + /* Try to fetch the MCU. */ + if (! (*cinfo->entropy->decode_mcu_discard_coef) (cinfo)) { + /* Suspension forced; update state counters and exit */ + coef->MCU_vert_offset = yoffset; + coef->MCU_ctr = MCU_col_num; + return JPEG_SUSPENDED; + } + } + /* Completed an MCU row, but perhaps not an iMCU row */ + coef->MCU_ctr = 0; + } + /* Completed the iMCU row, advance counters for next one */ + if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) { + start_iMCU_row(cinfo); + return JPEG_ROW_COMPLETED; + } + /* Completed the scan */ + (*cinfo->inputctl->finish_input_pass) (cinfo); + return JPEG_SCAN_COMPLETED; +} + +/* + * Same as consume_data, expect for saving the Huffman decode information + * - bitstream offset and DC coefficient to index. + */ + +METHODDEF(int) +consume_data_build_huffman_index_progressive (j_decompress_ptr cinfo, + huffman_index *index, int current_scan) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + JDIMENSION MCU_col_num; /* index of current MCU within row */ + int blkn, ci, xindex, yindex, yoffset; + JDIMENSION start_col; + JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN]; + JBLOCKROW buffer_ptr; + jpeg_component_info *compptr; + + int factor = 4; // maximum factor is 4. + for (ci = 0; ci < cinfo->comps_in_scan; ci++) + factor = jmin(factor, cinfo->cur_comp_info[ci]->h_samp_factor); + + int sample_size = index->MCU_sample_size * factor; + huffman_scan_header *scan_header = index->scan + current_scan; + scan_header->MCU_rows_per_iMCU_row = coef->MCU_rows_per_iMCU_row; + scan_header->MCUs_per_row = jdiv_round_up(cinfo->MCUs_per_row, sample_size); + scan_header->comps_in_scan = cinfo->comps_in_scan; + + size_t allocate_size = coef->MCU_rows_per_iMCU_row + * scan_header->MCUs_per_row * sizeof(huffman_offset_data); + scan_header->offset[cinfo->input_iMCU_row] = + (huffman_offset_data*)malloc(allocate_size); + index->mem_used += allocate_size; + + huffman_offset_data *offset_data = scan_header->offset[cinfo->input_iMCU_row]; + + /* Align the virtual buffers for the components used in this scan. */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + buffer[ci] = (*cinfo->mem->access_virt_barray) + ((j_common_ptr) cinfo, coef->whole_image[compptr->component_index], + 0, // Only need one row buffer + (JDIMENSION) compptr->v_samp_factor, TRUE); + } + /* Loop to process one whole iMCU row */ + for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row; + yoffset++) { + for (MCU_col_num = coef->MCU_ctr; MCU_col_num < cinfo->MCUs_per_row; + MCU_col_num++) { + /* For each MCU, we loop through different color components. + * Then, for each color component we will get a list of pointers to DCT + * blocks in the virtual buffer. + */ + blkn = 0; /* index of current DCT block within MCU */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + start_col = MCU_col_num * compptr->MCU_width; + /* Get the list of pointers to DCT blocks in + * the virtual buffer in a color component of the MCU. + */ + for (yindex = 0; yindex < compptr->MCU_height; yindex++) { + buffer_ptr = buffer[ci][yindex+yoffset] + start_col; + for (xindex = 0; xindex < compptr->MCU_width; xindex++) { + coef->MCU_buffer[blkn++] = buffer_ptr++; + if (cinfo->input_scan_number == 0) { + // need to do pre-zero by ourself. + jzero_far((void FAR *) coef->MCU_buffer[blkn-1], + (size_t) (SIZEOF(JBLOCK))); + } + } + } + } + // Record huffman bit offset + if (MCU_col_num % sample_size == 0) { + (*cinfo->entropy->get_huffman_decoder_configuration) + (cinfo, offset_data); + ++offset_data; + } + /* Try to fetch the MCU. */ + if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) { + /* Suspension forced; update state counters and exit */ + coef->MCU_vert_offset = yoffset; + coef->MCU_ctr = MCU_col_num; + return JPEG_SUSPENDED; + } + } + /* Completed an MCU row, but perhaps not an iMCU row */ + coef->MCU_ctr = 0; + } + (*cinfo->entropy->get_huffman_decoder_configuration) + (cinfo, &scan_header->prev_MCU_offset); + /* Completed the iMCU row, advance counters for next one */ + if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) { + start_iMCU_row(cinfo); + return JPEG_ROW_COMPLETED; + } + /* Completed the scan */ + (*cinfo->inputctl->finish_input_pass) (cinfo); + return JPEG_SCAN_COMPLETED; +} + +/* + * Decompress and return some data in the multi-pass case. + * Always attempts to emit one fully interleaved MCU row ("iMCU" row). + * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED. + * + * NB: output_buf contains a plane for each component in image. + */ + +METHODDEF(int) +decompress_data (j_decompress_ptr cinfo, JSAMPIMAGE output_buf) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1; + JDIMENSION block_num; + int ci, block_row, block_rows; + JBLOCKARRAY buffer; + JBLOCKROW buffer_ptr; + JSAMPARRAY output_ptr; + JDIMENSION output_col; + jpeg_component_info *compptr; + inverse_DCT_method_ptr inverse_DCT; + + /* Force some input to be done if we are getting ahead of the input. */ + while (cinfo->input_scan_number < cinfo->output_scan_number || + (cinfo->input_scan_number == cinfo->output_scan_number && + cinfo->input_iMCU_row <= cinfo->output_iMCU_row)) { + if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED) + return JPEG_SUSPENDED; + } + + /* OK, output from the virtual arrays. */ + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* Don't bother to IDCT an uninteresting component. */ + if (! compptr->component_needed) + continue; + /* Align the virtual buffer for this component. */ + buffer = (*cinfo->mem->access_virt_barray) + ((j_common_ptr) cinfo, coef->whole_image[ci], + cinfo->tile_decode ? 0 : cinfo->output_iMCU_row * compptr->v_samp_factor, + (JDIMENSION) compptr->v_samp_factor, FALSE); + /* Count non-dummy DCT block rows in this iMCU row. */ + if (cinfo->output_iMCU_row < last_iMCU_row) + block_rows = compptr->v_samp_factor; + else { + /* NB: can't use last_row_height here; it is input-side-dependent! */ + block_rows = (int) (compptr->height_in_blocks % compptr->v_samp_factor); + if (block_rows == 0) block_rows = compptr->v_samp_factor; + } + inverse_DCT = cinfo->idct->inverse_DCT[ci]; + output_ptr = output_buf[ci]; + int width_in_blocks = compptr->width_in_blocks; + int start_block = 0; +#if ANDROID_TILE_BASED_DECODE + if (cinfo->tile_decode) { + // width_in_blocks for a component depends on its h_samp_factor. + width_in_blocks = jmin(width_in_blocks, + (cinfo->coef->MCU_column_right_boundary - + cinfo->coef->MCU_column_left_boundary) * + compptr->h_samp_factor); + start_block = coef->pub.MCU_columns_to_skip * + compptr->h_samp_factor; + } +#endif + /* Loop over all DCT blocks to be processed. */ + for (block_row = 0; block_row < block_rows; block_row++) { + buffer_ptr = buffer[block_row]; + output_col = start_block * compptr->DCT_scaled_size; + buffer_ptr += start_block; + for (block_num = start_block; block_num < (unsigned int)width_in_blocks; block_num++) { + (*inverse_DCT) (cinfo, compptr, (JCOEFPTR) buffer_ptr, + output_ptr, output_col); + buffer_ptr++; + output_col += compptr->DCT_scaled_size; + } + output_ptr += compptr->DCT_scaled_size; + } + } + + if (++(cinfo->output_iMCU_row) < cinfo->total_iMCU_rows) + return JPEG_ROW_COMPLETED; + return JPEG_SCAN_COMPLETED; +} + +#endif /* D_MULTISCAN_FILES_SUPPORTED */ + + +#ifdef BLOCK_SMOOTHING_SUPPORTED + +/* + * This code applies interblock smoothing as described by section K.8 + * of the JPEG standard: the first 5 AC coefficients are estimated from + * the DC values of a DCT block and its 8 neighboring blocks. + * We apply smoothing only for progressive JPEG decoding, and only if + * the coefficients it can estimate are not yet known to full precision. + */ + +/* Natural-order array positions of the first 5 zigzag-order coefficients */ +#define Q01_POS 1 +#define Q10_POS 8 +#define Q20_POS 16 +#define Q11_POS 9 +#define Q02_POS 2 + +/* + * Determine whether block smoothing is applicable and safe. + * We also latch the current states of the coef_bits[] entries for the + * AC coefficients; otherwise, if the input side of the decompressor + * advances into a new scan, we might think the coefficients are known + * more accurately than they really are. + */ + +LOCAL(boolean) +smoothing_ok (j_decompress_ptr cinfo) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + boolean smoothing_useful = FALSE; + int ci, coefi; + jpeg_component_info *compptr; + JQUANT_TBL * qtable; + int * coef_bits; + int * coef_bits_latch; + + if (! cinfo->progressive_mode || cinfo->coef_bits == NULL) + return FALSE; + + /* Allocate latch area if not already done */ + if (coef->coef_bits_latch == NULL) + coef->coef_bits_latch = (int *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + cinfo->num_components * + (SAVED_COEFS * SIZEOF(int))); + coef_bits_latch = coef->coef_bits_latch; + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* All components' quantization values must already be latched. */ + if ((qtable = compptr->quant_table) == NULL) + return FALSE; + /* Verify DC & first 5 AC quantizers are nonzero to avoid zero-divide. */ + if (qtable->quantval[0] == 0 || + qtable->quantval[Q01_POS] == 0 || + qtable->quantval[Q10_POS] == 0 || + qtable->quantval[Q20_POS] == 0 || + qtable->quantval[Q11_POS] == 0 || + qtable->quantval[Q02_POS] == 0) + return FALSE; + /* DC values must be at least partly known for all components. */ + coef_bits = cinfo->coef_bits[ci]; + if (coef_bits[0] < 0) + return FALSE; + /* Block smoothing is helpful if some AC coefficients remain inaccurate. */ + for (coefi = 1; coefi <= 5; coefi++) { + coef_bits_latch[coefi] = coef_bits[coefi]; + if (coef_bits[coefi] != 0) + smoothing_useful = TRUE; + } + coef_bits_latch += SAVED_COEFS; + } + + return smoothing_useful; +} + + +/* + * Variant of decompress_data for use when doing block smoothing. + */ + +METHODDEF(int) +decompress_smooth_data (j_decompress_ptr cinfo, JSAMPIMAGE output_buf) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1; + JDIMENSION block_num, last_block_column; + int ci, block_row, block_rows, access_rows; + JBLOCKARRAY buffer; + JBLOCKROW buffer_ptr, prev_block_row, next_block_row; + JSAMPARRAY output_ptr; + JDIMENSION output_col; + jpeg_component_info *compptr; + inverse_DCT_method_ptr inverse_DCT; + boolean first_row, last_row; + JBLOCK workspace; + int *coef_bits; + JQUANT_TBL *quanttbl; + INT32 Q00,Q01,Q02,Q10,Q11,Q20, num; + int DC1,DC2,DC3,DC4,DC5,DC6,DC7,DC8,DC9; + int Al, pred; + + /* Force some input to be done if we are getting ahead of the input. */ + while (cinfo->input_scan_number <= cinfo->output_scan_number && + ! cinfo->inputctl->eoi_reached) { + if (cinfo->input_scan_number == cinfo->output_scan_number) { + /* If input is working on current scan, we ordinarily want it to + * have completed the current row. But if input scan is DC, + * we want it to keep one row ahead so that next block row's DC + * values are up to date. + */ + JDIMENSION delta = (cinfo->Ss == 0) ? 1 : 0; + if (cinfo->input_iMCU_row > cinfo->output_iMCU_row+delta) + break; + } + if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED) + return JPEG_SUSPENDED; + } + + /* OK, output from the virtual arrays. */ + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* Don't bother to IDCT an uninteresting component. */ + if (! compptr->component_needed) + continue; + /* Count non-dummy DCT block rows in this iMCU row. */ + if (cinfo->output_iMCU_row < last_iMCU_row) { + block_rows = compptr->v_samp_factor; + access_rows = block_rows * 2; /* this and next iMCU row */ + last_row = FALSE; + } else { + /* NB: can't use last_row_height here; it is input-side-dependent! */ + block_rows = (int) (compptr->height_in_blocks % compptr->v_samp_factor); + if (block_rows == 0) block_rows = compptr->v_samp_factor; + access_rows = block_rows; /* this iMCU row only */ + last_row = TRUE; + } + /* Align the virtual buffer for this component. */ + if (cinfo->output_iMCU_row > 0) { + access_rows += compptr->v_samp_factor; /* prior iMCU row too */ + buffer = (*cinfo->mem->access_virt_barray) + ((j_common_ptr) cinfo, coef->whole_image[ci], + (cinfo->output_iMCU_row - 1) * compptr->v_samp_factor, + (JDIMENSION) access_rows, FALSE); + buffer += compptr->v_samp_factor; /* point to current iMCU row */ + first_row = FALSE; + } else { + buffer = (*cinfo->mem->access_virt_barray) + ((j_common_ptr) cinfo, coef->whole_image[ci], + (JDIMENSION) 0, (JDIMENSION) access_rows, FALSE); + first_row = TRUE; + } + /* Fetch component-dependent info */ + coef_bits = coef->coef_bits_latch + (ci * SAVED_COEFS); + quanttbl = compptr->quant_table; + Q00 = quanttbl->quantval[0]; + Q01 = quanttbl->quantval[Q01_POS]; + Q10 = quanttbl->quantval[Q10_POS]; + Q20 = quanttbl->quantval[Q20_POS]; + Q11 = quanttbl->quantval[Q11_POS]; + Q02 = quanttbl->quantval[Q02_POS]; + inverse_DCT = cinfo->idct->inverse_DCT[ci]; + output_ptr = output_buf[ci]; + /* Loop over all DCT blocks to be processed. */ + for (block_row = 0; block_row < block_rows; block_row++) { + buffer_ptr = buffer[block_row]; + if (first_row && block_row == 0) + prev_block_row = buffer_ptr; + else + prev_block_row = buffer[block_row-1]; + if (last_row && block_row == block_rows-1) + next_block_row = buffer_ptr; + else + next_block_row = buffer[block_row+1]; + /* We fetch the surrounding DC values using a sliding-register approach. + * Initialize all nine here so as to do the right thing on narrow pics. + */ + DC1 = DC2 = DC3 = (int) prev_block_row[0][0]; + DC4 = DC5 = DC6 = (int) buffer_ptr[0][0]; + DC7 = DC8 = DC9 = (int) next_block_row[0][0]; + output_col = 0; + last_block_column = compptr->width_in_blocks - 1; + for (block_num = 0; block_num <= last_block_column; block_num++) { + /* Fetch current DCT block into workspace so we can modify it. */ + jcopy_block_row(buffer_ptr, (JBLOCKROW) workspace, (JDIMENSION) 1); + /* Update DC values */ + if (block_num < last_block_column) { + DC3 = (int) prev_block_row[1][0]; + DC6 = (int) buffer_ptr[1][0]; + DC9 = (int) next_block_row[1][0]; + } + /* Compute coefficient estimates per K.8. + * An estimate is applied only if coefficient is still zero, + * and is not known to be fully accurate. + */ + /* AC01 */ + if ((Al=coef_bits[1]) != 0 && workspace[1] == 0) { + num = 36 * Q00 * (DC4 - DC6); + if (num >= 0) { + pred = (int) (((Q01<<7) + num) / (Q01<<8)); + if (Al > 0 && pred >= (1<<Al)) + pred = (1<<Al)-1; + } else { + pred = (int) (((Q01<<7) - num) / (Q01<<8)); + if (Al > 0 && pred >= (1<<Al)) + pred = (1<<Al)-1; + pred = -pred; + } + workspace[1] = (JCOEF) pred; + } + /* AC10 */ + if ((Al=coef_bits[2]) != 0 && workspace[8] == 0) { + num = 36 * Q00 * (DC2 - DC8); + if (num >= 0) { + pred = (int) (((Q10<<7) + num) / (Q10<<8)); + if (Al > 0 && pred >= (1<<Al)) + pred = (1<<Al)-1; + } else { + pred = (int) (((Q10<<7) - num) / (Q10<<8)); + if (Al > 0 && pred >= (1<<Al)) + pred = (1<<Al)-1; + pred = -pred; + } + workspace[8] = (JCOEF) pred; + } + /* AC20 */ + if ((Al=coef_bits[3]) != 0 && workspace[16] == 0) { + num = 9 * Q00 * (DC2 + DC8 - 2*DC5); + if (num >= 0) { + pred = (int) (((Q20<<7) + num) / (Q20<<8)); + if (Al > 0 && pred >= (1<<Al)) + pred = (1<<Al)-1; + } else { + pred = (int) (((Q20<<7) - num) / (Q20<<8)); + if (Al > 0 && pred >= (1<<Al)) + pred = (1<<Al)-1; + pred = -pred; + } + workspace[16] = (JCOEF) pred; + } + /* AC11 */ + if ((Al=coef_bits[4]) != 0 && workspace[9] == 0) { + num = 5 * Q00 * (DC1 - DC3 - DC7 + DC9); + if (num >= 0) { + pred = (int) (((Q11<<7) + num) / (Q11<<8)); + if (Al > 0 && pred >= (1<<Al)) + pred = (1<<Al)-1; + } else { + pred = (int) (((Q11<<7) - num) / (Q11<<8)); + if (Al > 0 && pred >= (1<<Al)) + pred = (1<<Al)-1; + pred = -pred; + } + workspace[9] = (JCOEF) pred; + } + /* AC02 */ + if ((Al=coef_bits[5]) != 0 && workspace[2] == 0) { + num = 9 * Q00 * (DC4 + DC6 - 2*DC5); + if (num >= 0) { + pred = (int) (((Q02<<7) + num) / (Q02<<8)); + if (Al > 0 && pred >= (1<<Al)) + pred = (1<<Al)-1; + } else { + pred = (int) (((Q02<<7) - num) / (Q02<<8)); + if (Al > 0 && pred >= (1<<Al)) + pred = (1<<Al)-1; + pred = -pred; + } + workspace[2] = (JCOEF) pred; + } + /* OK, do the IDCT */ + (*inverse_DCT) (cinfo, compptr, (JCOEFPTR) workspace, + output_ptr, output_col); + /* Advance for next column */ + DC1 = DC2; DC2 = DC3; + DC4 = DC5; DC5 = DC6; + DC7 = DC8; DC8 = DC9; + buffer_ptr++, prev_block_row++, next_block_row++; + output_col += compptr->DCT_scaled_size; + } + output_ptr += compptr->DCT_scaled_size; + } + } + + if (++(cinfo->output_iMCU_row) < cinfo->total_iMCU_rows) + return JPEG_ROW_COMPLETED; + return JPEG_SCAN_COMPLETED; +} + +#endif /* BLOCK_SMOOTHING_SUPPORTED */ + + +/* + * Initialize coefficient buffer controller. + */ + +GLOBAL(void) +jinit_d_coef_controller (j_decompress_ptr cinfo, boolean need_full_buffer) +{ + my_coef_ptr coef; + + coef = (my_coef_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + SIZEOF(my_coef_controller)); + cinfo->coef = (struct jpeg_d_coef_controller *) coef; + coef->pub.start_input_pass = start_input_pass; + coef->pub.start_output_pass = start_output_pass; + coef->pub.column_left_boundary = 0; + coef->pub.column_right_boundary = 0; + coef->pub.MCU_columns_to_skip = 0; +#ifdef BLOCK_SMOOTHING_SUPPORTED + coef->coef_bits_latch = NULL; +#endif + +#ifdef ANDROID_TILE_BASED_DECODE + if (cinfo->tile_decode) { + if (cinfo->progressive_mode) { + /* Allocate one iMCU row virtual array, coef->whole_image[ci], + * for each color component, padded to a multiple of h_samp_factor + * DCT blocks in the horizontal direction. + */ + int ci, access_rows; + jpeg_component_info *compptr; + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + access_rows = compptr->v_samp_factor; + coef->whole_image[ci] = (*cinfo->mem->request_virt_barray) + ((j_common_ptr) cinfo, JPOOL_IMAGE, TRUE, + (JDIMENSION) jround_up((long) compptr->width_in_blocks, + (long) compptr->h_samp_factor), + (JDIMENSION) compptr->v_samp_factor, // one iMCU row + (JDIMENSION) access_rows); + } + coef->pub.consume_data_build_huffman_index = + consume_data_build_huffman_index_progressive; + coef->pub.consume_data = consume_data_multi_scan; + coef->pub.coef_arrays = coef->whole_image; /* link to virtual arrays */ + coef->pub.decompress_data = decompress_onepass; + } else { + /* We only need a single-MCU buffer. */ + JBLOCKROW buffer; + int i; + + buffer = (JBLOCKROW) + (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE, + D_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK)); + for (i = 0; i < D_MAX_BLOCKS_IN_MCU; i++) { + coef->MCU_buffer[i] = buffer + i; + } + coef->pub.consume_data_build_huffman_index = + consume_data_build_huffman_index_baseline; + coef->pub.consume_data = dummy_consume_data; + coef->pub.coef_arrays = NULL; /* flag for no virtual arrays */ + coef->pub.decompress_data = decompress_onepass; + } + return; + } +#endif + + /* Create the coefficient buffer. */ + if (need_full_buffer) { +#ifdef D_MULTISCAN_FILES_SUPPORTED + /* Allocate a full-image virtual array for each component, */ + /* padded to a multiple of samp_factor DCT blocks in each direction. */ + /* Note we ask for a pre-zeroed array. */ + int ci, access_rows; + jpeg_component_info *compptr; + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + access_rows = compptr->v_samp_factor; +#ifdef BLOCK_SMOOTHING_SUPPORTED + /* If block smoothing could be used, need a bigger window */ + if (cinfo->progressive_mode) + access_rows *= 3; +#endif + coef->whole_image[ci] = (*cinfo->mem->request_virt_barray) + ((j_common_ptr) cinfo, JPOOL_IMAGE, TRUE, + (JDIMENSION) jround_up((long) compptr->width_in_blocks, + (long) compptr->h_samp_factor), + (JDIMENSION) jround_up((long) compptr->height_in_blocks, + (long) compptr->v_samp_factor), + (JDIMENSION) access_rows); + } + coef->pub.consume_data = consume_data; + coef->pub.decompress_data = decompress_data; + coef->pub.coef_arrays = coef->whole_image; /* link to virtual arrays */ +#else + ERREXIT(cinfo, JERR_NOT_COMPILED); +#endif + } else { + /* We only need a single-MCU buffer. */ + JBLOCKROW buffer; + int i; + + buffer = (JBLOCKROW) + (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE, + D_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK)); + for (i = 0; i < D_MAX_BLOCKS_IN_MCU; i++) { + coef->MCU_buffer[i] = buffer + i; + } + coef->pub.consume_data = dummy_consume_data; + coef->pub.decompress_data = decompress_onepass; + coef->pub.coef_arrays = NULL; /* flag for no virtual arrays */ + } +} |