Image Processing: Valuable Information Extraction from Images !
Image Processing as we all know is a method of applying some operations on image either to enhance image quality or to extract some valuable information from image. Kind of signal processing where your input is image and output could either be a image or a feature associated with given input image.
Recently while working on a image processing use-case wherein the objective was to extract some particular type of information from some images of particular category/type. Now multiple options were in place to achieve this. Let's discuss about few of them in brief:
This use-case exposed different image processing algorithms and tools capabilities that could drive this use-case and solve the purpose. Use of OCR's like tesseract, GOCR, Ocrad, ocropus, Google OCR etc. was one option, other was use of built-in web tools like WebPlot Digitizer, Graph Maker - Plotly, OriginLab etc.. Also libraries like openCV also worked as image processor. Another option was writing some customised code in Matlab to extract data in natural images or Image Normalisation using languages like R programming also served the purpose.
Another important thing to focus on was the image format/type. For e.g. some figures could be in bitmap format whereas others could be in some vector format. In a bitmap format the image is stored by saving the colour code for each pixel. This means that information about overlapping data points is lost, because a pixel in a bitmap does not differentiate between different layers. However, in a vector format, information is stored on the shape and its position on the canvas, which is unrestricted to a specific pixel size, and information can be saved. As such, these images can be enlarged without loss of image quality. Moreover, the position of those shapes can be retraced in order to reconstruct data points in a figure. This can even be done when data points overlap, because unlike in the pixel format, overlapping shapes are stored alongside each other in a vector image.
There were many more challenges while extracting data from images like: Sometimes your data in image may not be in continuous Left to Right alignment . It may be slightly inclined at some angle then in that case two things matters: First was to bring crop program focus to right angle so that it extracts desired data/text from image. Second was the OCR or data extraction program may not be intelligent enough that it recognise each character of data/text in its original form then one need to deploy some spell check program or need some periodic human intervention to cross verify the data extracted from image.
Final step would be data merging after data extraction from bundle of images to particular format with desired image label across each set of extracted data. One of the ways to preserve all data together could be CSV or Excel workbook with different labels that summarise the extracted data from batch of images in Tabular format and similarly other formats could be used as per the requirement.