[Retracted] Study of Clustering Algorithms in Object Tracking and Image Segmentation

Algorithm 1

(1) The image to be segmented input to the network is defined as , and the prediction image result of the network is defined as .

All object classes of the dataset are defined as .

(2) Calculate the superresolution algorithm of image , and generate superpixel blocks with Mean Shift method, which are defined as

. s_j is one of the super-pixel block.

(3) For

All pixels in are defined as

get the prediction result of each pixel of

Count the number of pixels of each classes in , and the result is ®

(4) Change the prediction classes of each pixel in to

End for

(5) Update the output image to .

(6) Calculate the input image and the updated prediction result image to obtain the final prediction result .