simon987/libscan
1
0
Fork 0
mirror of https://github.com/simon987/libscan.git synced 2025-04-05 12:23:00 +00:00
Code Issues Projects Releases Wiki Activity

Save ebook renders as jpeg

Browse Source
This commit is contained in:
simon987 2020-07-17 20:18:21 -04:00
parent 0438c0e761
commit e7eca2c90a
5 changed files with 93 additions and 87 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

78
libscan/ebook/ebook.c
View File

@ -1,9 +1,10 @@
#include "ebook.h"
#include "../util.h"
#include <mupdf/fitz.h>
#include <pthread.h>
#include <tesseract/capi.h>
#include "../media/media.h"
#define MIN_OCR_SIZE 350
#define MIN_OCR_LEN 10
@ -71,32 +72,61 @@ int render_cover(scan_ebook_ctx_t *ctx, fz_context *fzctx, document_t *doc, fz_d
return -1;
}
fz_buffer *fzbuf = NULL;
fz_var(fzbuf);
fz_var(err);
fz_try(fzctx)
fzbuf = fz_new_buffer_from_pixmap_as_png(fzctx, pixmap, fz_default_color_params);
fz_catch(fzctx)
err = fzctx->error.errcode;
if (err == 0) {
unsigned char *tn_buf;
size_t tn_len = fz_buffer_storage(fzctx, fzbuf, &tn_buf);
APPEND_TN_META(doc, pixmap->x, pixmap->y)
ctx->store((char *) doc->uuid, sizeof(doc->uuid), (char *) tn_buf, tn_len);
}
fz_drop_buffer(fzctx, fzbuf);
fz_drop_pixmap(fzctx, pixmap);
fz_drop_page(fzctx, cover);
if (err != 0) {
CTX_LOG_WARNINGF(doc->filepath, "fz_new_buffer_from_pixmap_as_png() returned error code [%d] %s", err,
fzctx->error.message)
if (pixmap->n != 3) {
CTX_LOG_ERRORF(doc->filepath, "Got unexpected pixmap depth: %d", pixmap->n)
fz_drop_page(fzctx, cover);
fz_drop_pixmap(fzctx, pixmap);
return -1;
}
// RGB24 -> YUV420p
AVFrame *scaled_frame = av_frame_alloc();
struct SwsContext *sws_ctx= sws_getContext(
pixmap->w, pixmap->h, AV_PIX_FMT_RGB24,
pixmap->w, pixmap->h, AV_PIX_FMT_YUVJ420P,
SIST_SWS_ALGO, 0, 0, 0
);
int dst_buf_len = av_image_get_buffer_size(AV_PIX_FMT_YUV420P, pixmap->w, pixmap->h, 1);
uint8_t *dst_buf = (uint8_t *) av_malloc(dst_buf_len);
av_image_fill_arrays(scaled_frame->data, scaled_frame->linesize, dst_buf, AV_PIX_FMT_YUV420P, pixmap->w, pixmap->h, 1);
const uint8_t *in_data[1] = {pixmap->samples};
int in_line_size[1] = {3 * pixmap->w};
sws_scale(sws_ctx,
in_data, in_line_size,
0, pixmap->h,
scaled_frame->data, scaled_frame->linesize
);
scaled_frame->width = pixmap->w;
scaled_frame->height = pixmap->h;
scaled_frame->format = AV_PIX_FMT_YUV420P;
sws_freeContext(sws_ctx);
// YUV420p -> JPEG
AVCodecContext *jpeg_encoder = alloc_jpeg_encoder(pixmap->w, pixmap->h, 1.0f);
avcodec_send_frame(jpeg_encoder, scaled_frame);
AVPacket jpeg_packet;
av_init_packet(&jpeg_packet);
avcodec_receive_packet(jpeg_encoder, &jpeg_packet);
APPEND_TN_META(doc, pixmap->w, pixmap->h)
ctx->store((char *) doc->uuid, sizeof(doc->uuid), (char *) jpeg_packet.data, jpeg_packet.size);
av_packet_unref(&jpeg_packet);
av_free(*scaled_frame->data);
av_frame_free(&scaled_frame);
avcodec_free_context(&jpeg_encoder);
fz_drop_pixmap(fzctx, pixmap);
fz_drop_page(fzctx, cover);
return 0;
}

31
libscan/media/media.c
View File

@ -2,11 +2,6 @@
#include "../util.h"
#include "libavformat/avformat.h"
#include "libswscale/swscale.h"
#include "libswresample/swresample.h"
#include "libavcodec/avcodec.h"
#include "libavutil/imgutils.h"
#include <ctype.h>
@ -15,28 +10,6 @@
#define IS_VIDEO(fmt) (fmt->iformat->name && strcmp(fmt->iformat->name, "image2") != 0)
__always_inline
static AVCodecContext *alloc_jpeg_encoder(scan_media_ctx_t *ctx, int dstW, int dstH, float qscale) {
AVCodec *jpeg_codec = avcodec_find_encoder(AV_CODEC_ID_MJPEG);
AVCodecContext *jpeg = avcodec_alloc_context3(jpeg_codec);
jpeg->width = dstW;
jpeg->height = dstH;
jpeg->time_base.den = 1000000;
jpeg->time_base.num = 1;
jpeg->i_quant_factor = qscale;
jpeg->pix_fmt = AV_PIX_FMT_YUVJ420P;
int ret = avcodec_open2(jpeg, jpeg_codec, NULL);
if (ret != 0) {
CTX_LOG_WARNINGF("media.c", "Could not open jpeg encoder: %s!\n", av_err2str(ret));
return NULL;
}
return jpeg;
}
__always_inline
AVFrame *scale_frame(const AVCodecContext *decoder, const AVFrame *frame, int size) {
@ -362,7 +335,7 @@ void parse_media_format_ctx(scan_media_ctx_t *ctx, AVFormatContext *pFormatCtx,
}
// Encode frame to jpeg
AVCodecContext *jpeg_encoder = alloc_jpeg_encoder(ctx, scaled_frame->width, scaled_frame->height,
AVCodecContext *jpeg_encoder = alloc_jpeg_encoder(scaled_frame->width, scaled_frame->height,
ctx->tn_qscale);
avcodec_send_frame(jpeg_encoder, scaled_frame);
@ -602,7 +575,7 @@ int store_image_thumbnail(scan_media_ctx_t *ctx, void* buf, size_t buf_len, docu
}
// Encode frame to jpeg
AVCodecContext *jpeg_encoder = alloc_jpeg_encoder(ctx, scaled_frame->width, scaled_frame->height, ctx->tn_qscale);
AVCodecContext *jpeg_encoder = alloc_jpeg_encoder(scaled_frame->width, scaled_frame->height, ctx->tn_qscale);
avcodec_send_frame(jpeg_encoder, scaled_frame);
AVPacket jpeg_packet;

28
libscan/media/media.h
View File

@ -4,6 +4,12 @@
#include "../scan.h"
#include "libavformat/avformat.h"
#include "libswscale/swscale.h"
#include "libswresample/swresample.h"
#include "libavcodec/avcodec.h"
#include "libavutil/imgutils.h"
typedef struct {
log_callback_t log;
logf_callback_t logf;
@ -14,6 +20,28 @@ typedef struct {
long max_media_buffer;
} scan_media_ctx_t;
__always_inline
static AVCodecContext *alloc_jpeg_encoder(int w, int h, float qscale) {
AVCodec *jpeg_codec = avcodec_find_encoder(AV_CODEC_ID_MJPEG);
AVCodecContext *jpeg = avcodec_alloc_context3(jpeg_codec);
jpeg->width = w;
jpeg->height = h;
jpeg->time_base.den = 1000000;
jpeg->time_base.num = 1;
jpeg->i_quant_factor = qscale;
jpeg->pix_fmt = AV_PIX_FMT_YUVJ420P;
int ret = avcodec_open2(jpeg, jpeg_codec, NULL);
if (ret != 0) {
return NULL;
}
return jpeg;
}
void parse_media(scan_media_ctx_t *ctx, vfile_t *f, document_t *doc);
void init_media();

39
libscan/raw/raw.c
View File

@ -1,37 +1,8 @@
#include "raw.h"
#include <libraw/libraw.h>
#include <errno.h>
#include <unistd.h>
#include "../media/media.h"
#include "libswscale/swscale.h"
#include "libswresample/swresample.h"
#include "libavcodec/avcodec.h"
#include "libavutil/imgutils.h"
__always_inline
static AVCodecContext *alloc_jpeg_encoder(scan_raw_ctx_t *ctx, int dstW, int dstH, float qscale) {
AVCodec *jpeg_codec = avcodec_find_encoder(AV_CODEC_ID_MJPEG);
AVCodecContext *jpeg = avcodec_alloc_context3(jpeg_codec);
jpeg->width = dstW;
jpeg->height = dstH;
jpeg->time_base.den = 1000000;
jpeg->time_base.num = 1;
jpeg->i_quant_factor = qscale;
jpeg->pix_fmt = AV_PIX_FMT_YUVJ420P;
int ret = avcodec_open2(jpeg, jpeg_codec, NULL);
if (ret != 0) {
CTX_LOG_WARNINGF("raw.c", "Could not open jpeg encoder: %s!\n", av_err2str(ret))
return NULL;
}
return jpeg;
}
#define MIN_SIZE 32
@ -152,11 +123,11 @@ void parse_raw(scan_raw_ctx_t *ctx, vfile_t *f, document_t *doc) {
av_image_fill_arrays(scaled_frame->data, scaled_frame->linesize, dst_buf, AV_PIX_FMT_YUV420P, dstW, dstH, 1);
const uint8_t *inData[1] = {img->data};
int inLinesize[1] = {3 * img->width};
const uint8_t *in_data[1] = {img->data};
int in_line_size[1] = {3 * img->width};
sws_scale(sws_ctx,
inData, inLinesize,
in_data, in_line_size,
0, img->height,
scaled_frame->data, scaled_frame->linesize
);
@ -167,7 +138,7 @@ void parse_raw(scan_raw_ctx_t *ctx, vfile_t *f, document_t *doc) {
sws_freeContext(sws_ctx);
AVCodecContext *jpeg_encoder = alloc_jpeg_encoder(ctx, scaled_frame->width, scaled_frame->height, 1.0f);
AVCodecContext *jpeg_encoder = alloc_jpeg_encoder(scaled_frame->width, scaled_frame->height, 1.0f);
avcodec_send_frame(jpeg_encoder, scaled_frame);
AVPacket jpeg_packet;

4
test/main.cpp
View File

@ -174,12 +174,16 @@ TEST(Ebook, CandlePdf) {
document_t doc;
load_doc_file("libscan-test-files/test_files/ebook/General_-_Candle_Making.pdf", &f, &doc);
size_t size_before = store_size;
parse_ebook(&ebook_500_ctx, &f, "application/pdf", &doc);
ASSERT_STREQ(get_meta(&doc, MetaTitle)->str_val, "Microsoft Word - A531 Candlemaking-01.doc");
ASSERT_STREQ(get_meta(&doc, MetaAuthor)->str_val, "Dafydd Prichard");
ASSERT_NEAR(strlen(get_meta(&doc, MetaContent)->str_val), 500, 4);
ASSERT_NE(get_meta(&doc, MetaContent)->str_val[0], ' ');
ASSERT_NE(size_before, store_size);
cleanup(&doc, &f);
}