| Summary: | The use of digital image-based visual information has gained a lot of attention due to its flexibility and easy to be manipulated. However, digital images that have been used as basic input to an application system are frequently contaminated by noise. One of the most common types of noise found in digital images is impulse noise. Therefore, two new types of filters based on switching scheme are proposed as impulse noise removal. Both filters namely Dual Sliding Statistics Switching Median filter (MPSDG) and Adroit Hybrid Switching Median filter (MPHM) are two-stage filters which consist of noise detection and noise filtering stages. In the MPSDG method, the filtering process begins with noise detection stage; whereby the statistics of a localized detection window of sorted and non-sorted orders are simultaneously processed. Subsequently, the median of absolute difference obtained from the statistics of both windows will be used to classify the existence of the noise pixel. Next in the filtering stage, the pixels that have been classified as noise will be restored while the noise-free pixels will be left unchanged. Meanwhile in the MPHM technique, the statistics in the local window detectors are arranged in ascending order and the difference of each data with the median value will be calculated. Then all of the obtained difference values will be processed in the next steps and threshold values will be used to distinguish between the noise pixels and noise-free pixels. Once the noise detection process is completed, all the noise pixels are replaced with the estimated median value in the next filtering stage. By using these processing techniques, the proposed filters are not only able to remove the impulse noise, but are also able to maintain the structure and shape of the object in the image. The simulation results also show that the MPSDG and MPHM filters outperform other existing conventional filters in the literature, in terms of qualitative and quantitative assessments.
|
|---|