9 August 2024 in International Journal of Computer Vision
13 March 2017 in International Journal of Computer Vision
1 December 2017 in Pattern Recognition
8 May 2019 in IEEE Transactions on Image Processing
Spatial misalignment caused by variations in poses and viewpoints is one of the most critical issues that hinder the performance improvement in existing person re-identification (Re-ID) algorithms. Although it is straightforward to explore correspondence learning algorithms for alignment, online learning is intractable for negative pairs due to the intrinsic visual difference between negative pairs and efficiency concern. To address this problem, in this paper, we present a robust and efficient graph correspondence transfer (REGCT) approach for explicit spatial alignment in Re-ID. Specifically, we propose the off-line correspondence learning and on-line correspondence transfer framework. During training, patch-wise correspondences between positive training pairs are established via graph matching. By exploiting both spatial and visual contexts of human appearance in graph matching, meaningful semantic correspondences can be obtained. During testing, the off-line learned patch-wise correspondence templates are transferred to test pairs with similar pose-pair configurations for local feature distance calculation. To enhance the robustness of correspondence transfer, we design a novel pose context descriptor to accurately model human body configurations, and present an approach to measure the similarity between a pair of pose context descriptors. Meanwhile, to improve testing efficiency, we propose a correspondence template ensemble method using the voting mechanism, which significantly reduces the amount of patch-wise matchings involved in distance calculation. With aforementioned strategies, the REGCT model can effectively and efficiently handle the spatial misalignment problem in Re-ID. Extensive experiments on five challenging benchmarks, including VIPeR, Road, PRID450S, 3DPES and CUHK01, evidence the superior performance of REGCT over other state-of-the-art approaches.
8 July 2020 in IEEE Access
Blind image deblurring, composed of estimating blur kernel and non-blind deconvolution, is an extremely ill-posed problem. However, previous deblurring methods still cannot solve delta kernel or noise problem well and avoid ringing artifacts in restored image, because the reliable kernel estimation and image restoration could not be given from information-deficiency input without using natural image priors. In this work, we find that the blur process changes the distribution of the image gradient, and thus attempt to use the priori knowledge for guiding blind deblurring. For the sake of convenience of modeling, we come up with a simplified approximate formulation of the image gradient distribution prior, thus the restoration model using it can be solved by the method of iteratively reweighted least squares (IRLS). We also concentrate on how to optimize the models and develop an algorithm based on gradient prior and image structure: first, we computed image structure based on the total variation model and adaptive weight strategy and then estimated the strong edges from it. Those strong edges (structure) that have a possible adverse effect on blur kernel estimation can be removed. Next, we followed an alternate iterative framework to obtain high-quality blur kernel estimation by estimating blur kernel from strong structure and then restoring a latent image guided by the formulated clear image priors. Finally, we proposed a non-blind deconvolution method based on fitted multi-order gradient priors as regularization to restore the latent image. In addition, we analyze the effectiveness of the knowledge-driven prior in image deblurring and demonstrate that it can favor clear images over blurred ones and restrain ringing artifacts effectively. Extensive experiments verify the superiority of the proposed method over state-of-the-art algorithms compared, both qualitatively and quantitatively.