Blotch Detection Based on Texture Matching and Adaptive Multi-threshold

Abstract

Blotch is a typical artifact in old films and the detection of them is an important step in film restoration. The existing simplified rank-ordered difference detector achieves higher detection rates by reducing the value of threshold. However, the corresponding higher number of false alarms is undesirable. To maximize the ratio between correct detections and false alarms, this paper proposes an improved blotch detector based on adaptive multithreshold. According to different objects of blotches, the proposed detector can achieve the most appropriate threshold by convergence confinement. Meanwhile, texture matching is introduced to avoid the possible deviation caused by motion vector estimation in the regions with blotches. Performance evaluation is taken to the image sequences with both real blotches and artificially corrupted ones. The experimental results indicate higher correct detection rates and fewer false alarms simultaneously.

Motion compensation under varying illumination for wavelet-based blotch detection in old movies

“…However, in case of a low correlation between the blocks, a variable size block strategy could be performed. The most important property of the PhC method is that it is robust to illumination variations thanks to the normalization performed in Equation (4). Finally, the motion compensated frame ( −1) is the result of compensating each block…”

“…Some of these techniques operate in the spatial domain such as the Spike Detection Index (SDI) detector [1], the Simplified Rank Ordered Differences (SROD) method [2] and detectors based on AR models [1], [2]. Recently, improved blotch detection approaches exploiting judiciously the SROD have been proposed [3], [4] in order to reduce the false alarm level. A statistical strategy within the a contrario framework was recently proposed in [5] in order to decide, for each pixel, whether the differences between its neighborhood and the neighborhood of its corresponding pixel in the previous (respectively the next) frame are significant.…”

Sea moving target detection method based on mixed multi-Gaussian difference model

“…The algorithm of moving object detection based on empty salient feature has achieved good results in accuracy and robustness. Liu W. et al 5 used the three-frame difference method to extract the object contour from the video image, and used the Gaussian mixture background difference between adaptive learning rate to update the model quickly, increase the iteration times of the background model, and eliminate the "ghost shadow" caused by the object motion. On this basis, the learning rate was adjusted based on the difference between the target pixel and the adjacent eight pixels in the current frame and the background model, which is modified adaptively.…”

“…It needs to obtain the moving object in the image quickly in limited time, and ensure the integrity of the internal details and contour of the target. 5,6 According to different data acquisition methods, the commonly used sea-surface moving target monitoring systems can be divided into satellite remote sensing system, sea-surface monitoring radar system, infrared imaging system and visible light video monitoring system. 7, 8 Satellite remote sensing system has long processed period and low resolution and then it can't realize real-time monitoring.…”

Detection of Blotch and Scratch in Video Based on Video Decomposition

“…But the difficulty of determining these thresholds restrains the practical effectiveness of this method. Therefore, a simplified ROD detector (S-ROD) that uses only one threshold has been proposed in [5] and [46]. Methods in [6] and [45] further improve ROD's performance by using more robust motion estimation and extending ROD to five motion compensated (MC) frames.…”