Traditional image signal processors (ISPs) are primarily designed and optimized to improve the image quality perceived by humans. However, optimal perceptual image quality does not always translate into optimal performance for computer vision applications. We propose a set of methods, which we collectively call VisionISP, to repurpose the ISP for machine consumption. VisionISP significantly reduces data transmission needs by reducing the bit-depth and resolution while preserving the relevant information. The blocks in VisionISP are simple, content-aware, and trainable. Experimental results show that VisionISP boosts the performance of a subsequent computer vision system trained to detect objects in an autonomous driving setting. The results demonstrate the potential and the practicality of VisionISP for computer vision applications.
Fig. 1. Comparison of ISP output images with different tunings on IMX260 ISO400. From left to right: Not-tuned, Hand-tuned by image quality expert, Auto-tuned.
ABSTRACTImage Signal Processor (ISP) comprises of various blocks to reconstruct image sensor raw data to final image consumed by human visual system or computer vision applications. Each block typically has many tuning parameters due to the complexity of the operation. These need to be hand tuned by Image Quality (IQ) experts, which takes considerable amount of time. In this paper, we present an automatic IQ tuning using nonlinear optimization and automatic reference generation algorithms. The proposed method can produce high quality IQ in minutes as compared with weeks of handtuned results by IQ experts. In addition, the proposed method can work with any algorithms without being aware of their specific implementation. It was found successful on multiple different processing blocks such as noise reduction, demosaic, and sharpening.
Camera modules have become more popular in consumer electronics and office products. As a consequence, people have many opportunities to use a camera-based device to record a hardcopy document in their daily lives. However, it is easy to let undesired shading into the captured document image through the camera. Sometimes, this non-uniformity may degrade the readability of the contents. In order to mitigate this artifact, some solutions have been developed. But most of them are only suitable for particular types of documents.In this paper, we introduce a content-independent and shape-independent method that will lessen the shading effects in captured document images. We want to reconstruct the image such that the result will look like a document image captured under a uniform lighting source. Our method utilizes the 3D depth map of the document surface and a look-up table strategy. We will first discuss the model and the assumptions that we used for the approach. Then, the process of creating and utilizing the look-up table will be described in the paper.We implement this algorithm with our prototype 3D scanner, which also uses a camera module to capture a 2D image of the object. Some experimental results will be presented to show the effectiveness of our method. Both flat and curved surface document examples will be included.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.