研究目的
The problem of single image dehazing consists in recovering an undegraded image by processing degraded images that were captured in a scattering medium. It should be noted that this problem is challenging because several physical factors need to be estimated, such as the depth distribution function of the scene, the concentration density of suspended particles in the medium, and the magnitude of environmental light, among others.
研究成果
The proposed GP-based methodology for designing estimators of the medium transmission function for the restoration of images degraded by haze has shown superior performance compared to state-of-the-art methods. The evolved estimators effectively remove haze without introducing noticeable overprocessing effects or edge artifacts, making them suitable for real-time applications. The performance improvements are statistically significant in many test cases.
研究不足
The main drawback of existing single image dehazing methods is that they introduce undesirable edge artifacts and overprocessing effects to the restored images. The proposed GP-based methodology aims to minimize these effects but may still face challenges in scenarios with extremely hazy conditions or complex scene geometries.
1：Experimental Design and Method Selection:
The methodology involves using genetic programming to evolve computer programs that estimate the medium transmission function of hazy scenes. These programs are composed of basic mathematical operators and are optimized with respect to the mean-absolute-error.
2：Sample Selection and Data Sources:
A set of synthetic hazy images and their ground-truth transmission functions are used as the training set. These images represent various challenging situations for image dehazing, such as abrupt depth changes and highly textured objects.
3：List of Experimental Equipment and Materials:
The experiments are conducted using a Desktop computer with a 3.4 GHz Intel Core i7 processor and 16 GB of RAM memory. The methods are implemented in Python, utilizing the DEAP and neat-GP libraries.
4：4 GHz Intel Core i7 processor and 16 GB of RAM memory. The methods are implemented in Python, utilizing the DEAP and neat-GP libraries.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: The GP algorithm is executed in four different configurations, considering two GP-variants and two approaches for the extraction of local image information. The performance of the evolved estimators is evaluated in terms of objective metrics (MAE and PSNR) and subjective visual criteria.
5：Data Analysis Methods:
The performance of the evolved estimators is compared with state-of-the-art methods using the MAE and PSNR metrics. Statistical significance tests are performed to validate the results.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
