Research On Image Compression Based On Image Segmentation

Image compression is one of the most important branch of image processing. With thedevelopment of the technology, the requirement in compression rate and visual effects getshigher. Image compression based on image segmention needs to segment images first andthen compress each region depending on degree of interest of the region. Image compressionbased on image segmention can adapt people’s requirement.The paper introduced some techniques about image segmentation and imagecompression. Wavelet theory makes great contribution to image compression. The paperdescribed wavelet theory in detail. Image compression based on image segmention needs tosegment images first. In this paper, the watershed algorithm based on mathematicalmorphology was researched and improved, wavelet was used to transform each segment, andthen an adaptive morphological dilation image coding algorithm was used to represent thewavelet data. The main contents are as follows:Due to the noisy points and local extreme minimum values, watershed transform has theproblem of over-segmentation. The paper proposed a new algorithm combined watershedtransform and region growing algorithm. At present there are lots of segmentation evaluationcriterions. The region homogeneity and region contrast are the classical and commonevaluation criterions. A good segmentation result means intra-region similar property (regionhomogeneity is good) and inter-region dissimilar property (region contrast is big). In thispaper, a new evaluation criterion was proposed which was based on Shannon entropy bycombining the two properties together.Generally wavelet transform was used on rectangular region, a transform rule wasproposed in this paper which makes it possible to use wavelet transform on irregular region.The statistical properties of wavelet coefficients of image data was studied. According to thestatistical properties, an adaptive morphological dilation image coding algorithm waspresented.