International Congress on Image and Signal Processing 

About: International Congress on Image and Signal Processing is an academic conference. The conference publishes majorly in the area(s): Feature extraction & Image segmentation. Over the lifetime, 5485 publications have been published by the conference receiving 21947 citations.

An available database for the research of finger vein recognition

Abstract: Finger vein biometric has received considerable attentions in recent years. However, there is no approved and benchmark finger vein database for researchers to evaluate their algorithms. Furthermore, few public finger vein databases are available online. Aiming to support a benchmark database, in this paper, we introduce a representative finger vein database captured by a portable device, which is named MMCBNU_6000. Our research is novel in four aspects. First, MMCBNU_6000 is established with participation of 100 volunteers, coming from 20 countries. It contains images acquired from different persons with different skin colors. Second, statistical information of the nationality, age, gender, and blood type is recorded for further analysis on finger vein images. Third, similar to the real application, influences from translation, rotation, scale, uneven illumination, scattering, collection posture, finger tissue and finger pressure are taken into account in the imaging process. Fourth, according to the evaluation of average image gray value, image contrast and entropy on the images from the available databases, the acquired images in MMCBNU_6000 have comparable image quality.

Improvement of canny algorithm based on pavement edge detection

Abstract: In this paper we introduce an improved Canny edge detection algorithm and an edge preservation filtering procedure for pavement edge detection applications. Data of pavement images were randomly selected to test this algorithm. There are some problems of Canny operator, unable to detect the weak edge and distinguish the grayscale with little change, the detected edge uncontinuous. Based on these defects, the paper mainly uses the Mallat wavelet transform to reinforce the weak edge of input images, quadratic optimization of genetic algorithm to get a proper threshold in self-adapting standard during Canny algorithm steps. With the base of Canny operator and the improvement, the paper builds a new model, which satisfies the need of pavement edge detection real-time. Computer simulations show that the improved algorithm can make up for the disadvantages of Canny algorithm, detect edges of pavement images effectively, and is a less time-consuming process. Particularly, it has been shown that the presented algorithm can not only eliminate noises effectively but also protect unclear edges.

Convolutional neural networks based transfer learning for diabetic retinopathy fundus image classification

Abstract: Convolutional Neural Networks (CNNs) have gained remarkable success in computer vision, which is mostly owe to their ability that enables learning rich image representations from large-scale annotated data. In the field of medical image analysis, large amounts of annotated data may be not always available. The number of acquired ground-truth data is sometimes insufficient to train the CNNs without overfitting and convergence issues from scratch. Hence application of the deep CNNs is a challenge in medical imaging domain. However, transfer learning techniques are shown to provide solutions for this challenge. In this paper, our target task is to implement diabetic retinopathy fundus image classification using CNNs based transfer learning. Experiments are performed on 1014 and 1200 fundus images from two publicly available DR1 and MESSIDOR datasets. In order to complete the target task, we carry out experiments using three different methods: 1) fine-tuning all network layers of each of different pre-trained CNN models; 2) fine-tuning a pre-trained CNN model in a layer-wise manner; 3) using pre-trained CNN models to extract features from fundus images, and then training support vector machines using these features. Experimental results show that convolutional neural networks based transfer learning can achieve better classification results in our task with small datasets (target domain), by taking advantage of knowledge learned from other related tasks with larger datasets (source domain). Transfer learning is a promising technique that promotes the use of deep CNNs in medical field with limited amounts of data.

Fog Removal from Color Images using Contrast Limited Adaptive Histogram Equalization

Abstract: The images degraded by fog suffer from poor contrast. In order to remove fog effect, a Contrast Limited Adaptive Histogram Equalization (CLAHE)-based method is presented in this paper. This method establishes a maximum value to clip the histogram and redistributes the clipped pixels equally to each gray-level. It can limit the noise while enhancing the image contrast. In our method, firstly, the original image is converted from RGB to HSI. Secondly, the intensity component of the HSI image is processed by CLAHE. Finally, the HSI image is converted back to RGB image. To evaluate the effectiveness of the proposed method, we experiment with a color image degraded by fog and apply the edge detection to the image. The results show that our method is effective in comparison with traditional methods.

Image recognition of plant diseases based on backpropagation networks

Abstract: To achieve automatic diagnosis of plant diseases and improve the image recognition accuracy of plant diseases, two kinds of grape diseases (grape downy mildew and grape powdery mildew) and two kinds of wheat diseases (wheat stripe rust and wheat leaf rust) were selected as research objects, and the image recognition of the diseases was conducted based on image processing and pattern recognition. After image preprocessing including image compression, image cropping and image denoising, K_means clustering algorithm was used to segment the disease images, and then 21 color features, 4 shape features and 25 texture features were extracted from the images. Backpropagation (BP) networks were used as the classifiers to identify grape diseases and wheat diseases, respectively. The results showed that identification of the diseases could be effectively achieved using BP networks. While the dimensions of the feature data were not reduced by using principal component analysis (PCA), the optimal recognition results for grape diseases were obtained as the fitting accuracy and the prediction accuracy were both 100%, and that for wheat diseases were obtained as the fitting accuracy and the prediction accuracy were both 100%. While the dimensions of the feature data were reduced by using PCA, the optimal recognition result for grape diseases was obtained as the fitting accuracy was 100% and the prediction accuracy was 97.14%, and that for wheat diseases was obtained as the fitting accuracy and the prediction accuracy were both 100%.