Color Images Of Hypercomplex Overall Processing Theory And Applied Research

Hypercomplexs encode the three channels of a color image as a whole vector for further processing. Comparing with the traditional methods, the hypercomplex processing methods can better describe the intrinsic relationships of different color channels and break new ground to color image processing.Firstly, fast algorithms of hypercomplex Fourier transform and hypercomplex correlation are proposed in this dissertation, which calculate each part Fourier transform respectively with FFT algorithm by separating a hypercomplex into real and imaginary parts. When reverting the corresponding unit imaginary operators and combining the results of the real and imaginary parts Fourier transforms, we can get the hypercomplex results. The mathematical analytical results show that our algorithms are easier to implement and have less computing complexity than the known algorithms.Next, we propose a novel hypercomplex domain content-based watermarking algorithm for color host images, which embedding the watermark data into proper hypercomplex frequency components obtained through the hypercomplex Fourier thansform of the host images. We explore the prerequisite to ensure that the watermarked host image can also be transferred in three channels, which can be satisfied by our proposed technique. The analytical results show that the image distortions from the watermark are spreaded into the three channels of the host image by hypercomplex inverse Fourier transform. So the watermarks embedded by our proposed algorithm have both good imperceptibility and security. At the same time, a cross redundancy embedding rule is proposed to improve watermarking robustness. Furthermore, based on different mask properties of the human vision system about texture, edge and luminance of the color host images, the selected frequency spectras are weighted adaptively. The superiority of the proposed algorithm is verified by comparing our experimental results with the known ones in the literatures.Moreover, a color object tracking algorithm in hypercomplex space is proposed by treating color image object pixel as a whole vector for processing. We set up a color object model in hypercomplex domain, which provides integrated color characteristics of the object. This algorithm gives small weights to pixels farther from the object center to improve the robustness to the interference from the background. To reduce the influence of outlier, we calculate the distance-weighted hypercomplex robust correlation by fast hypercomplex correlation algorithm to identify the moving object's location. And we use image feature recognition to acquire the moving object's scale, which aims to overcome the change of the object appearance. Thus we realize a color object tracking algorithm which is robust to occlusions and interferences from the background. Video experimental results verify the effectiveness of the proposed algorithm.