A new unsupervised pseudo-siamese network with two filling strategies for image denoising and quality enhancement

Digital image noise may be introduced during acquisition, transmission, or processing and affects readability and image processing effectiveness. The accuracy of established image processing techniques, such as segmentation, recognition, and edge detection, is adversely impacted by noise. There exists an extensive body of work which focuses on circumventing such issues through digital image enhancement and noise reduction, but this work is limited by a number of constraints including the application of non-adaptive parameters, potential loss of edge detail information, and (with supervised approaches) a requirement for clean, labeled, training data. This paper, developed on the principle of Noise2Void, presents a new unsupervised learning approach incorporating a pseudo-siamese network. Our method enables image denoising without the need for clean images or paired noise images, instead requiring only noise images. Two independent branches of the network utilize different filling strategies, namely zero filling and adjacent pixel filling. Then, the network employs a loss function to improve the similarity of the results in the two branches. We also modify the Efficient Channel Attention module to extract more diverse features and improve performance on the basis of global average pooling. Experimental results show that compared with traditional methods, the pseudo-siamese network has a greater improvement on the ADNI dataset in terms of quantitative and qualitative evaluation. Our method therefore has practical utility in cases where clean images are difficult to obtain.

Pengujian Modifikasi Kernel Konvolusi Untuk Penajaman Dan Penghalusan Citra Berwarna

<p>Suatu citra digital terkadang memiliki kualitas yang kurang baik, misalnya terdapat noise, terlalu tajam, kurang tajam, kabur, terlalu gelap, terlalu terang, dan sebagainya. Untuk itu diperlukan perbaikan kualitas citra digital (image enhancement) sebagai proses awal pengolahan citra (image preprocessing) agar citra dapat digunakan untuk proses selanjutnya, seperti pengenalan pola (patern recognition). Konvolusi merupakan salah satu teknik perbaikan kualitas citra yang digunakan untuk memanipulasi piksel-piksel dalam citra, sehingga diperoleh citra yang memiliki sifat dan informasi yang diinginkan. Penelitian ini bertujuan untuk melakukan eksperimen terhadap kernel/tapis konvolusi dengan beberapa pengujian untuk penajaman dan penghalusan citra berwarna. Hasil eksperimen/pengujian menunjukkan bahwa tingkat ketajaman suatu citra hasil konvolusi bergantung pada nilai tiap elemen dan ukuran matriks kernel. Semakin besar angka dan ukuran matriks yang digunakan maka semakin tinggi pula tingkat ketajaman yang dihasilkan. Pada penghalusan citra, tingkat kehalusan bergantung pada ukuran matriks kernel. Semakin besar ukuran matriks kernel maka semakin besar tingkat kehalusannya (semakin blur). Dari penelitian ini diharapkan dapat membantu mengetahui kernel yang tepat untuk memperbaiki kualitas suatu citra, sehingga dapat digunakan untuk proses lebih lanjut. <br />Kata kunci: konvolusi, kernel, citra digital</p>

Quantitative evaluation of digital-image enhancement during heads-up surgery

Image-processing is an advantage of heads-up surgery and expected to facilitate ophthalmic surgeries. To evaluate image-processing quantitatively, we analyzed the surgical images of twenty eyes that underwent vitrectomy with internal limiting membrane (ILM) peeling assisted by Brilliant Blue G (BBG). Still images of the peeling procedure were obtained from the surgical video, and the color difference was calculated between two adjacent spots inside and outside the ILM-peeling contour, i.e., without and with BBG staining, respectively. The color differences were compared between the two settings with and without image-processing, delivered by an algorithm to enhance the color and contrast. Color differences were calculated using two methods: the Euclidean distance based on RGB values (RGB distance) and the Delta-E00 formula provided by the International Commission on Illumination. In five cases, minimum light intensities required to recognize the contour of ILM-peeling were compared during surgeries between the two settings with and without enhancement. Image-processing increased the mean color difference significantly (P < 0.001) from 15.47 and 4.49 to 34.03 and 8.00, respectively, for the RGB distance and Delta-E00. The minimum light intensity was reduced from 15 to 5 on average by image-enhancement. These results showed image-processing enhances color differences and reduces light intensities during vitrectomy.

USING DIGITAL TERRESTRIAL PHOTOGRAMMETRY IN CULTURAL HERITAGE CASE STUDY: THE GREAT WALL OF GORGAN IN IRAN

An assessment of the influence of using digital terrestrial photogrammetry for surveying and documentation of cultural objects is presented in this paper. The approaches including digital image enhancement, digital rectification and restitution, feature extraction for the creation of a three-dimension geographical information system model from the photogrammetric record and the computer visualisation of cultural monuments. Manual three-dimension processing of terrestrial images using analogue photogrammetric procedures is slow, can register little information and has limited application and cannot be re-examined if the information desired is not directly presented. In addition, it is a very time-consuming task and requires the skill of qualified personnel. It seems there is a need for an environment-based information system with the ability to display precise and measurable imagery for use of the architectural and archaeological information system by integrating digital photogrammetry and AutoCAD facilities as applicable to support the reconstruction of many cultural heritage places. Architectural structuring and guidelines can be used to develop invaluable historical monuments in Iran such as the Great Wall of Gorgan. While the research is in completion, the output of the Great Wall of Gorgan can be documented and recorded in the world.