Research On Key Technique Of ROI Detection For Static Image

Image region of interest(ROI for short) detection is one of the hotspots in image processing technology. With the help of region of interest detection technology, image data compression can be guided, observers are helped understand the images, it also applies to display image on small screen. Based on the existing research, this paper studies and improves the region of interest detection algorithm based on visual features.For static images, the region of interest detection method using color difference and entropy is simple and effective, but the result is not very good when using a single approach. An idea that adding the factor of area into the criterion is presented in this paper, in order to achieve the balance between image compression rate and information reservation. Meanwhile, for portrait photographs, the face detection technology is proposed to help increase the accuracy of the region of interest detection.For the method of combining two factors of color difference and entropy to detect region of interest, two solutions are presented in this paper. One is to get intersection and union of the two results. The intersection result is the most interested region for observers in the image, while the union result can reduce the rate of improper detection and increase the accuracy. Another method is to use the weighting sum of two factors as criteria of region of interest detection. In order to determine the weighting parameters, the use of statistical sample and image feature extraction method is presented in this paper. Based on the experimental results of sample images, pictures can be classified into groups. Then calculate the average value of image features for each group, identify the contact between weighting parameters and image features. Experimental results show that there are some contacts between weighting parameters and the features of color histogram and Hu moment invariants.