ABSTRAKSI: Suatu citra digital merupakan salah satu bentuk citra yang populer karena memiliki kemudahan dalam pengambilan, pengiriman, maupun pemrosesannya. Suatu citra digital sering mengalami kerusakan atau penurunan kualitas citra yang disebabkan oleh kondisi lingkungan yang kurang baik pada saat pengambilan sebuah citra digital seperti adanya debu. Untuk itu dibutuhkan sebuah filter noise untuk meningkatkan kualitas citra yang telah mengalami kerusakan sehingga diperoleh citra yang kualitasnya sama atau mendekati kualitas citra digital sebelum terkena noise.
Pada tugas akhir ini dibangun sebuah sistem yang menguji kombinasi pendeteksi dan filter noise dalam mengurangi noise pada sebuah citra digital. Noise yang diuji adalah impulsive noise yang membangkitkan noise dengan mengubah nilai intensitas sebuah pixel menjadi intensitas maksimum atau intensitas minimum pada citra grayscale. Metode pendeteksian impulsive noise yang digunakan adalah Boundary Resetting Boundary Discriminative Noise Detection (BRBDND) dan Signal Dependent Rank Order Mean (SDROM). Sedangkan metode filter yang digunakan adalah Adaptive Minimum Mean Square Error Filter (Adaptive MMSE Filtering). BRBDND mendeteksi impulsive noise dengan mengelompokkan pixel yang diuji kedalam low-density cluster, middle cluster, dan high intensity cluster dimana pixel yang diuji termasuk impulsive noise apabila berada pada kelompok low-density cluster dan high intensity cluster.
Metode SDROM melakukan pendeteksian impulsive noise dengan menentukan 4 nilai ambang (threshold) yang dibandingkan dengan perbedaan nilai (sign difference) antara pixel tetangga dan pixel pusat pada matrik mask 3x3. Jika terdapat nilai sign difference yang melebihi nilai threshold, maka pixel yang diuji merupakan impulsive noise. Metode Adaptive MMSE Filtering melakukan filter dengan merata-ratakan nilai pixel tetangga dari pixel yang dikenai proses filter. Ukuran mask yang digunakan untuk melakukan proses filter adalah 3x3, 5x5, dan 7x7 dimana dilakukan dianalisis pengaruh ukuran mask terhadap hasil filter kombinasi pendeteksi impulsive noise dengan Adaptive MMSE Filtering berdasarkan nilai parameter objektif Peak Signal to Noise Ratio (PSNR).
Kombinasi SDROM+Adaptive MMSE Filtering dan kombinasi BRBDND+Adaptive MMSE Filtering mampu melakukan kombinasi untuk meningkatkan kualitas citra yang terkena impulsive noise dimana kombinasi BRBDND+Adaptive MMSE Filtering mampu meningkatkan nilai PSNR citra ternoise hingga mencapai selisih rata-rata sebesar 10.3658 dB. Sedangkan pada kombinasi SDROM+Adaptive MMSE Filtering mampu meningkatkan nilai PSNR citra ter-noise hingga mencapai selisih rata-rata sebesar 8.4209 dB.
Kata Kunci : Adaptive MMSE Filtering, Boundary Resetting BoundaryABSTRACT: A digital imagery is a form of popular imagery for having ease in collecting, sending, and processing. A digital image are often damaged or image quality degradation caused by unfavorable environmental conditions at the time of making a digital image such as the presence of dust. That requires a noise filter to improve the quality of image that has been damaged, so the image quality will be obtained the same or close to the quality of the digital image before affected by noise.
In this final task will be built a system that test the combination of detector and noise filter to reduce noise in a digital image. Noise to be tested is impulsive noise which generate noise by changing the intensity of a pixel value becomes the maximum intensity or minimum intensity in grayscale image. Impulsive noise detection method used was Boundary Resetting Boundary Discriminative Noise Detection (BRBDND) and Signal Dependent Rank Order Mean (SDROM). While the filter method to be used is the Adaptive MMSE Filtering (Adaptive MMSE Filtering). Boundary Resetting Boundary Discriminative Noise Detection detected impulsive noise pixels are tested and classified into low-density cluster, middle cluster, and high-intensity cluster in where the pixel being examined including the noise if they are in the group of low-density cluster and high-intensity cluster.
Method of Signal Dependent Rank Order Mean make detection of impulsive noise by determining the fourth threshold value to be compared with the sign difference between neighboring pixels and the central pixel in the matrix mask of 3x3. If there is a signed difference value that exceeds the threshold value, then the tested pixel is noise. Method of Adaptive MMSE Filtering make the filter with uniformly averaged pixel values of neighboring pixels to be subjected to filter process. Size mask that was used to filter process is a 3x3, 5x5, and 7x7 which will analyze the influence of mask size on the result of a combination of detector noise filter with the Adaptive MMSE Filtering based on an objective parameter value of Peak Signal to Noise Ratio (PSNR).
The combination of SDROM+Adaptive MMSE Filtering and combination of BRBDND+Adaptive MMSE Filtering able to combine to increase the image quality is affected by impulsive noise, where the combination BRBDND+Adaptive MMSE Filtering can improve image PSNR value was noise until it reaches an average difference of 10.3658 dB. While the combination SDROM+Adaptive MMSE Filtering can improve image PSNR value was noise until it reaches an average difference of 8.4209 dB.
Keyword: Adaptive MMSE Filtering, Boundary Resetting Boundary