To better achieve image correction effect, an image correction method based on corner point detection is proposed. In image preprocessing, firstly, image equalization is achieved based on the contrast limited adaptive histgram equalization to avoid the problems caused by illumination and suppress noise while maintaining details, and then adaptive threshold segmentation is performed using the OTSU to obtain the binarized image. In the corner point detection stage, the contours of the binarized image are extracted firstly and the closed contours are filled to avoid the independent contours from affecting the accuracy of region growth, then the center point of the image is used as the seed pixel for region growth, and finally the four corner points are calculated based on the linear contours of the growth region and Hough theory, where the accurate region growth result can avoid the influence of the background on the corner point detection. In the correction stage, the perspective matrix is calculated by the four corner points, and the image is corrected by the perspective transformation. The experiments show that the proposed method can accurately find the corner points of document images and achieve efficient correction.

An image correction method based on corner point detection

SRGAN applies generative adversarial networks to super-resolution reconstruction, and the advent of this algorithm greatly facilitates image reconstruction based on deep learning. It introduces batch normalization for stable training, and although batch normalization can facilitate training, it destroys the original contrast information of the image and affects the quality of image reconstruction. For this problem, this paper proposes an super-resolution algorithm based on improved SRGAN, which removes the batch normalization in the residual block of the generative network to avoid its bad influence on the reconstructed image and improve the quality of the reconstructed image for better. In addition, the total variation is introduced in the loss function of the discriminative model to constrain the gradient change, which serves to stabilize the network training and accelerate the convergence. Experiments of the proposed super-resolution algorithm on a large number of public datasets show that this method achieves better results under various evaluation metrics compared with other similar algorithms.

Super-resolution algorithm using an improved generative adversarial network

In this paper, a fast defect detection method for oranges is studied from the perspective of machine vision. Firstly, in order to eliminate the interference of orange stems in defect detection, a shape-based template matching method is used to extract the orange stems region and the average pixels of the image are used to fill the region. Secondly, contrast enhancement and edge extraction are performed on the image with orange stems eliminated, and the background is removed by combining morphological processing and filling techniques to obtain the orange region. Finally, by carrying out channel separation on the orange region image, the OSTU threshold segmentation is performed on its red channel image according to the color characteristics of the orange and its defects, and the extraction and contour fitting of the defects are completed by combining area features. The experiments show that the proposed method can quickly and accurately achieve the defect detection of oranges, and then further provide the core detection technology for some automatic orange sorting. 

A fast defect detection method for orange

In recent years, significant progress has been made in image super-resolution (SR) methods based on convolutional neural networks. However, most of them do not fully utilize multi-scale feature correspondence in the image SR process, resulting in blurred and artifact detail restoration, especially for SR tasks with larger scaling factors (i.e. $\times 4$ and $\times 8$ ). A multi-scale feature-enhanced SR network (MFENet) is proposed to solve the problems mentioned above. Specifically, the super-resolution method is enhanced by utilizing multi-scale feature correspondence in two ways. On the one hand, we propose an effective multi-scale non-local attention module and it can fully exploit a large number of low-level features at different scales of a single low-resolution (LR) feature map by exhaustively evaluating the correlations in local feature, in-scale non-local feature, and cross-scale non-local feature. On the other hand, the mapping from high-resolution (HR) to LR images is explored by a feedback branch we introduced, that provides an additional constraint for images reconstruction. It can penalize incorrect SR predictions to achieve better SR performance. Experimental results on 5 benchmark datasets show that our method outperforms state-of-the-art methods in terms of both subjective visual quality and objective quantitative metrics.

Yikai Gu

Papers

An image correction method based on corner point detection

Super-resolution algorithm using an improved generative adversarial network

A fast defect detection method for orange

Learning Enriched Features for Image Super Resolution