Methods and systems for segmenting multiple documents from a single input image
The present disclosure discloses methods and systems (for example multi-function devices) for segmenting multiple documents from a single input image. The multi-functional device includes a controller unit having a boundary extraction module to process an input image having multiple documents. The processing is performed using a computer implemented sum of absolute difference (SAD) based edge detection technique, to identify potential edge pixels and the region of interest pixels of the plurality of document images. Based on the identified potential edge pixels, perimeter boundaries surrounding each of the plurality of document images are determined. The determined perimeter boundaries and ROI pixels (Image mask for each document) in the input image are then segmented to create separate pages or files for each of the documents present within the input image for the purpose of printing separate files.
1. A method for segmenting multiple documents from a single input image, comprising:
receiving the input image comprising the multiple documents;
identifying potential edge pixels of the multiple documents using a computer implemented sum of absolute difference (SAD) based edge detection technique, wherein the identifying comprising:
segmenting each color separation of the input image into equal sized overlapped sub-images for each pixel, and for each sub-image;
determining a maximum pixel value for each color separation in the each sub-image;
computing the SAD value between the maximum pixel value and pixel values in the sub-image for each color separation;
determining average SAD value of the SAD values computed in the sub-image for each color separation; and
based on the average SAD value, identifying the pixels of the sub-image as the potential edge pixels;
based on the identified potential edge pixels, determining perimeter boundaries surrounding a region of interest corresponding to each of the multiple documents using morphological operations;
generating an image mask representing the region of interest for each of the multiple documents;
based on determined perimeter boundaries and generated image mask representing the region of interest for each document, segmenting the multiple documents from the input image; and
printing separate document files for each of the segmented multiple documents.
2. The method as claimed in claim 1 , wherein the identifying potential edge pixels of the multiple documents comprising separating the input image in Red (R), Green (G), Blue (B) color spaces.
3. The method as claimed in claim 1 , wherein identifying potential edge pixels of the sub-image as the potential edge pixels when the average SAD value is above a defined threshold value.
4. The method as claimed in claim 3 , wherein the defined threshold value is 19.
5. The method as claimed in claim 1 , wherein determining of perimeter boundaries comprising:
performing closing operation on each of the multiple documents present in the input image;
performing hole-filling operation to fill holes with pixel values that represent colors that are similar to the color of background pixels that are in the vicinity of each of the holes; and
following closing and hole filling operation, counting the number of documents present in the input image.
6. The method as claimed in claim 1 , further comprising:
extracting individual images for each of the segmented documents using selective segmented mask image technique;
compressing the extracted images corresponding to the multiple segmented documents; and
storing the extracted images in separate files on one of a storage device and a memory device.
7. The method as claimed in claim 1 , further comprising:
performing Hough transform on the edges of each of the multiple documents to determine non-zero value of orientation of the individual documents; and
rotating the individual documents for the non-zero value of the orientation determined by the Hough transform.
8. A multi-functional device for segmenting multiple documents from a single input image, comprising:
a controller unit comprising:
a processor;
a boundary extraction module, coupled to the processor, to:
receive an input image having multiple documents;
identify potential edge pixels of the multiple documents using a computer implemented sum of absolute difference (SAD) based edge detection technique, wherein the technique comprises:
segment each color separation of the input image into equal sized overlapped sub-images for each pixel, and for each sub-image;
determine a maximum pixel value for each color separation in each sub-image;
compute the SAD value between the maximum pixel value and pixel values in the sub-image for each color separation;
determine average SAD value of the SAD values computed in the sub-image for each color separation; and
based on the average SAD value, identify the pixels of the sub-image as the potential edge pixels; and
based on the identified potential edge pixels, determine perimeter boundaries surrounding a region of interest corresponding to each of the multiple documents using morphological operations; and a segmentation module, coupled to the processor, to:
generate an image mask representing the region of interest for each of the multiple documents;
segment the multiple documents in the input image based on the determined perimeter boundaries and the generated image mask representing the region of interest for each of the multiple documents; and
storing separate document image files for each of the segmented multiple documents.
9. The multi-function device as claimed in claim 8 , further comprising a flatbed scanning surface upon which the multiple documents are placed for receiving the input image.
10. The multi-function device as claimed in claim 8 , wherein the boundary extraction module identifies the potential edge pixels of the sub-image as the potential edge pixels when the average SAD value is above a defined threshold value.
11. The multi-function device as claimed in claim 10 , wherein the defined threshold value is 19.
12. The multi-function device as claimed in claim 8 , wherein the boundary extraction module determines the perimeter boundaries by:
performing a closing operation on each of the multiple documents;
performing hole-filling operation to fill holes with pixel values that represent colors that are similar to the color of background pixels that are in the vicinity of each of the holes; and
following the closing and hole filling operations, counting the number of the multiple documents present in the input image.
13. The multi-function device as claimed in claim 8 , wherein the segmentation module creates the separate image files by:
extracting individually images for each of the segmented documents using selective segmented mask image technique;
compressing the extracted images of each of the segmented documents; and
storing the extracted images in the separate image files on one of a storage device and a memory device.
14. The multi-function device as claimed in claim 8 , further comprising a Hough transform module, coupled to the processor, to:
perform Hough transform on the edges of the multiple documents to determine non-zero value of orientation of individual documents; and
rotate the individual documents for the non-zero value of the orientation determined by the Hough transform.
15. A non-transitory computer-readable medium comprising instructions executable by a processing resource to:
receive an input image comprising multiple documents;
identify potential edge pixels of the multiple documents using a computer implemented sum of absolute difference (SAD) based edge detection technique, wherein the technique comprises:
segment each color separation of the input image into equal sized overlapped sub-images for each pixel, and for each sub-image,
determine a maximum pixel value for each color separation;
compute the SAD value between the maximum pixel value and pixel values in the sub-image for each color separation;
determine average SAD value of the SAD values computed in the sub-image for each color separation; and
based on the average SAD value, identify the pixels of the sub-image as the potential edge pixels;
based on the identified potential edge pixels, determine perimeter boundaries surrounding a region of interest corresponding to each of the multiple documents using one or more morphological operations;
generate an image mask representing the region of interest for each of the multiple documents;
segment the multiple documents in the input image based on the determined perimeter boundaries and generated image mask representing the region of interest for each of the multiple documents; and
print separate document files for each of the segmented multiple documents.
16. The non-transitory computer-readable medium as claimed in claim 15 comprising instructions executable by the processing resource to identify the potential edge pixels of the sub-image as the potential edge pixels when the average SAD value is above a defined threshold value.
17. The non-transitory computer-readable medium as claimed in claim 16 , wherein the defined threshold value is 19.
18. The non-transitory computer-readable medium as claimed in claim 15 comprising instructions executable by the processing resource to:
perform closing operation on each of the multiple documents;
perform hole-filling operation to fill holes with pixel values that represent colors that are similar to the color of background pixels that are in the vicinity of each of the holes; and
follow closing and hole filling operation, count the number of multiple documents present within the input image.
19. The non-transitory computer-readable medium as claimed in claim 15 comprising instructions executable by the processing resource to:
extract individually images for each of the multiple segmented documents using selective segmented mask image technique;
compress the extracted images of each of the multiple segmented documents; and
store the extracted images in separate image files on one of a storage device and a memory device.
20. The non-transitory computer-readable medium as claimed in claim 15 comprising instructions executable by the processing resource to:
perform Hough transform on the edges of the multiple documents to determine non-zero value of orientation of individual documents; and
rotate the individual documents for the non-zero value of the orientation determined by the Hough transform.