Text detection in continuous tone image segments

A computer-implemented method of compressing a continuous tone region of an image containing text, the method comprising:

• on a pixel-by-pixel basis, classifying pixels of the region as being continuous tone or text;
• forming a palettized image region corresponding to the continuous tone region wherein pixels classified as being continuous tone are assigned a background color value;
• compressively encoding the palettized image region using a first compression process suited to palettized image compression;
• and
• compressively encoding the continuous tone region using a second compression process suited to continuous tone image compression.

A computer-implemented method of detecting pixels of textual content in a continuous tone region of an image, the method comprising:

• identifying pixels in the continuous tone region that are non-smooth relative to the values of their neighboring pixels;
• identifying contiguous groups of the non-smooth pixels that have a same-value;
• and
• determining whether pixel counts of the identified contiguous groups exceed a size threshold;
• and
• classifying pixels that form an identified contiguous group whose pixel count exceeds the size threshold as textual content.

A computer-readable medium having computer-executable instructions stored thereon for performing a method of separating pixels of textual content in a continuous tone region of an image, the method comprising:

• separating pixels corresponding to textual content within the continuous tone region;
• encoding the separated textual content pixels using a first compressor suited to palettized image compression;
• and
• encoding the continuous tone region using a second compressor suited to continuous-tone image compression;
• forming a compressed image data stream containing the encoded separated textual content pixels data and the encoded continuous tone region data.

A computer system comprising:

• a text segmenter operating to process pixels in a continuous tone region of an image to detect contiguous strands of same-valued pixels forming text, and to generate a separate image map containing the detected text pixels;
• a palettized image encoder operating to compressively encode the separate image map of detected text pixels using a palettized image compression process;
• and
• a continuous tone image encoder operating to compressively encode the continuous tone region using a continuous tone image compression process;
• and
• a multiplexer operating to combine the encoded separate image map and continuous tone region into a compressed output data.

A computer-implemented method of decompressing a compressed data stream representing a mixed-mode image with text and continuous tone content, the compressed data stream containing a continuous tone segment in which continuous tone content for a region of the image is encoded using a compression type suitable for continuous tone content, the compressed data stream also containing a separate textual content segment in which text pixels detected as contiguous same-value pixel strands exceeding a size threshold from the image region are encoded using a compression type suitable for palettized image content, the method comprising:

• decoding the continuous tone segment to reconstruct continuous tone content of the image region;
• separately decoding the separate textual content segment to reconstruct the separated text pixels;
• and
• superimposing the separated text pixels onto the continuous tone content.

A mixed-mode image decoder for decoding a compressed data stream representing a mixed-mode image with text and continuous tone content, the compressed data stream containing a continuous tone segment in which continuous tone content for a region of the image is encoded using a compression type suitable for continuous tone content, the compressed data stream also containing a separate textual content segment in which text pixels detected as contiguous same-value pixel strands exceeding a size threshold from the image region are encoded using a compression type suitable for palettized image content, the decoder comprising:

• a demultiplexer operating to separate the separately encoded continuous tone and textual content segments from the compressed data stream;
• a palettized image decoder operating to decode the separate textual content segment to reconstruct the text pixels;
• a continuous tone image decoder operating to decode the continuous tone segment to reconstruct continuous tone content of the image region;
• and
• an image processor operating to superimpose the decoded text pixels over the reconstructed continuous tone content.

A computer-readable medium having computer-executable instructions stored thereon for decompressing a compressed data stream representing a mixed-mode image with text and continuous tone content, the compressed data stream containing a continuous tone segment in which continuous tone content for a region of the image is encoded using a compression type suitable for continuous tone content, the compressed data stream also containing a separate textual content segment in which text pixels detected as contiguous same-value pixel strands exceeding a size threshold from the image region are encoded using a compression type suitable for palettized image content, the method comprising:

• decoding the continuous tone segment to reconstruct continuous tone content of the image region;
• separately decoding the separate textual content segment to reconstruct the separated text pixels;
• and
• superimposing the separated text pixels onto the continuous tone content.