Download presentation
Presentation is loading. Please wait.
Published byMadison Dalton Modified over 9 years ago
1
DESIGNING AND MAKING OF NOISE REDUCTION APPLICATION USING IMAGE MORPHOLOGY by : Dionisius Kristal / 26406061
2
Preliminary 0 The image result from the digital camera is not suit with the expected result. 0 The camera that produces image with a little noise is very expensive. 0 Some people develop algorithms to eliminate the noise (called noise reduction).
3
Previous Work 0 In 1986, Sternberg introduced the idea of noise reduction by repeatedly opening and closing with an increasing structuring elements size. 0 Song and Delp in 1990, discovered a technique they called "generalized morphological filter". However, use of structuring elements must be precise so that results can be maximized.
4
Theory 0 Image Processing Digital image processing is a discipline that studies matters relating to improvement of image quality. 0 Morphology Morphology is the science of form and structure. It is about regions or shapes, how they can be changed and counted, and how their areas can be evaluated.
5
Theory 0 Dilation to expand/grow
6
Theory 0 Erosion to reduce/shrink
7
Theory Dilation Erosion EXPANDEDREDUCED
8
Theory 0 Opening (Erosion, then Dilation)
9
Theory 0 Closing (Dilation, then Erosion)
10
Theory 0 Peak Signal to Noise Ratio Peak^2 is the peak pixel value between two images. RMSE is square root of MSE. The Sum of (Original Image – Result Image)^2/(width*height) ^2
11
Theory What is Noise Reduction? Noise Reduction is program/system that has ability to reduce the (image) noise. Original ImageResult Image
12
Theory Noise Reduction using Mathematical Morphology? 0 “Noisy” image = clean image + noise 0 Segment into features and noise (the residual image). 0 Residual image = the difference between an original image and smoothed version. 0 The features from residual image will be added back to the smoothed image. 0 The results is an image whose edges and other one dimensional features are as sharp as the original one, but has smooth regions between them.
13
Original Image
14
Result Image
15
How it works? 0 The process is divided by 2 processes : smoothing process, and detail recovery. 0 Smoothing process is OCCO filtering (morphologycal opening-closing-closing-opening) 0 Detail recovery process is TOPBOT filtering (Tophat and Bothat) where Tophat is a positive residual image and Bothat is a negative residual image. 0 The output of TOPBOT filtering is Tophat and Bothat accumulation, where in this case, Tophat and Bothat accumulation is clean from noise.
16
How it works? The final result : “Clean” image = Tophat Accumulation + OCCO Image – Bothat Accumulation
17
How it works? OCCO image Tophat Accumulation Bothat Accumulation “Clean” image = Tophat Accumulation + OCCO Image – Bothat Accumulation
18
How it works? OCCO image
19
How it works? Final Result
20
Flowchart (main) OCCO filtering Tophat filtering Bothat filtering Final Summary
21
Flowchart OCCO O=OPEN(I) OC=CLOSE(O) C=CLOSE(I) OCCO image = ½ (OC+CO) CO=OPEN(C)
22
Flowchart TOPBOT
23
System Design and Application 0 3 modules of the program are : Load Module opening an image Noise Reduction Module processing the noisy image Save Image Module saving the image processing result.
24
Experiment High ISO images are shot with ISO 400 and ISO 800 Then those images are being processed with the program and Photoshop The image results will be compared with Low ISO image and there will be a number that indicate the PSNR score
25
Experiment There are 6 categories of image : 0 Image with small particle object (Kopi.jpg, Komputer.jpg, Tombol.jpg) 0 Image with bright object and dark background (Pasir.jpg, Beras.jpg) 0 Image with letter (Box.jpg, Majalah.jpg, Notes.jpg, Koran.jpg) 0 Image with certain pattern (Lemari.jpg, Jeans.jpg, Batik.jpg) 0 Face Image (Wajah1.jpg, Wajah2.jpg) 0 Image with Noise Generator added on it
26
Experiment Table (Kodak M1033 ISO 400) ISOGambarPerulanganIndeksPSNR MIC NRPSNR Photoshop 400 Box.jpg (Gambar 5.42) 1Gambar 5.3825,43 32,01 2Gambar 5.3925,13 3Gambar 5.4024,93 Kopi.jpg (Gambar 5.44) 1Gambar 5.4124,06 31,15 2Gambar 5.4223,89 3Gambar 5.4323,84 Majalah.jpg (Gambar 5.46) 1Gambar 5.4427,33 25,21 2Gambar 5.4527,39 3Gambar 5.4627,33 Notes.jpg (Gambar 5.48) 1Gambar 5.4716,51 28,77 2Gambar 5.4816,35 3Gambar 5.4916,29 Pasir.jpg (Gambar 5.50) 1Gambar 5.4724,3 29,43 2Gambar 5.4824,09 3Gambar 5.4924,02 Tombol.jpg (Gambar 5.52) 1Gambar 5.3836,65 36,56 2Gambar 5.3936,32 3Gambar 5.4036,15 Komputer.jp g (Gambar 5.54) 1Gambar 5.4132,39 32,68 2Gambar 5.4231,82 3Gambar 5.4331,53 Koran.jpg (Gambar 5.56) 1Gambar 5.4421,35 23,28 2Gambar 5.4520,92 3Gambar 5.4620,72 Lemari.jpg (Gambar 5.58) 1Gambar 5.4733,14 33,7 2Gambar 5.4832,19 3Gambar 5.4931,69 Beras.jpg (Gambar 5.60) 1Gambar 5.4730,1 31,02 2Gambar 5.4829,34 3Gambar 5.4928,88 Batik.jpg (Gambar 5.62) 1Gambar 5.4727,1 27,76 2Gambar 5.4826,03 3Gambar 5.4925,49 Jeans.jpg (Gambar 5.64) 1Gambar 5.4720,25 24,61 2Gambar 5.4819,34 3Gambar 5.4919,27
27
Experiment Table (Kodak M1033 ISO 800) ISOGambarPerulanganIndeksPSNR MIC NRPSNR Photoshop 800 Box.jpg (Gambar 5.43) 1Gambar 5.3826,68 31,25 2Gambar 5.3926,28 3Gambar 5.4026,01 Kopi.jpg (Gambar 5.45) 1Gambar 5.4128,96 32,44 2Gambar 5.4228,65 3Gambar 5.4328,57 Majalah.jpg (Gambar 5.47) 1Gambar 5.4424,62 23,69 2Gambar 5.4524,63 3Gambar 5.4624,57 Notes.jpg (Gambar 5.49) 1Gambar 5.4716,10 27,89 2Gambar 5.4815,93 3Gambar 5.4915,89 Pasir.jpg (Gambar 5.51) 1Gambar 5.4725,15 26,91 2Gambar 5.4824,94 3Gambar 5.4924,87 Tombol.jpg (Gambar 5.53) 1Gambar 5.3835,68 36,23 2Gambar 5.3934,62 3Gambar 5.4034,35 Komputer.jp g (Gambar 5.55) 1Gambar 5.4129,87 30,17 2Gambar 5.4229,43 3Gambar 5.4329,17 Koran.jpg (Gambar 5.57) 1Gambar 5.4420,39 21,32 2Gambar 5.4520,21 3Gambar 5.4620,1 Lemari.jpg (Gambar 5.59) 1Gambar 5.4733,05 33,23 2Gambar 5.4832,33 3Gambar 5.4931,94 Beras.jpg (Gambar 5.61) 1Gambar 5.4726,67 26,8 2Gambar 5.4826,56 3Gambar 5.4926,34 Batik.jpg (Gambar 5.63) 1Gambar 5.4723,66 24,04 2Gambar 5.4823,14 3Gambar 5.4922,82 Jeans.jpg (Gambar 5.65) 1Gambar 5.4719,86 21,5 2Gambar 5.4819,34 3Gambar 5.4919,27
28
Experiment Table (Nikon D90 ISO 400) ISOGambarPerulanganIndeksPSNR MIC NRPSNR Photoshop 400 Box.jpg (Gambar 5.66) 1Gambar 5.3824,38 27,54 2Gambar 5.3924,44 3Gambar 5.4024,22 Kopi.jpg (Gambar 5.68) 1Gambar 5.4121,57 24,54 2Gambar 5.4229,54 3Gambar 5.4330,44 Majalah.jpg (Gambar 5.70) 1Gambar 5.4430,53 31,78 2Gambar 5.4529,85 3Gambar 5.4629,56 Notes.jpg (Gambar 5.72) 1Gambar 5.4717,96 28,63 2Gambar 5.4817,5 3Gambar 5.4917,7 Pasir.jpg (Gambar 5.74) 1Gambar 5.4725,81 29,95 2Gambar 5.4822,37 3Gambar 5.4920,64 Tombol.jpg (Gambar 5.76) 1Gambar 5.3837,95 37,96 2Gambar 5.3937,18 3Gambar 5.4036,91 Komputer.j pg (Gambar 5.78) 1Gambar 5.4133,33 33,71 2Gambar 5.4232,77 3Gambar 5.4332,42 Koran.jpg (Gambar 5.80) 1Gambar 5.4424,2 25,72 2Gambar 5.4523,57 3Gambar 5.4623,69 Lemari.jpg (Gambar 5.82) 1Gambar 5.4737,49 37,45 2Gambar 5.4836,67 3Gambar 5.4935,45 Beras.jpg (Gambar 5.84) 1Gambar 5.4730,74 35,8 2Gambar 5.4829,1 3Gambar 5.4928,63 Batik.jpg (Gambar 5.86) 1Gambar 5.4722,31 22,38 2Gambar 5.4821,97 3Gambar 5.4921,8 Jeans.jpg (Gambar 5.88) 1Gambar 5.4717,15 16,25 2Gambar 5.4817,29 3Gambar 5.4917,28
29
Experiment Table (Nikon D90 ISO 800) ISOGambarPerulanganIndeksPSNR MIC NR PSNR Photoshop 800 Box.jpg (Gambar 5.67) 1Gambar 5.3824,39 27.86 2Gambar 5.3924,08 3Gambar 5.4023,85 Kopi.jpg (Gambar 5.69) 1Gambar 5.4130,77 22.48 2Gambar 5.4230,53 3Gambar 5.4330,44 Majalah.jpg (Gambar 5.71) 1Gambar 5.4425,95 26.07 2Gambar 5.4525,79 3Gambar 5.4625,07 Notes.jpg (Gambar 5.73) 1Gambar 5.4717,67 28.25 2Gambar 5.4817,5 3Gambar 5.4917,39 Pasir.jpg (Gambar 5.75) 1Gambar 5.4725,3 31.04 2Gambar 5.4825,08 3Gambar 5.4924,90 Tombol.jpg (Gambar 5.77) 1Gambar 5.3839,01 39,15 2Gambar 5.3938,12 3Gambar 5.4037,75 Komputer.j pg (Gambar 5.79) 1Gambar 5.4130,83 30,93 2Gambar 5.4230,56 3Gambar 5.4330,33 Koran.jpg (Gambar 5.81) 1Gambar 5.4422,21 22,85 2Gambar 5.4521,97 3Gambar 5.4621,90 Lemari.jpg (Gambar 5.83) 1Gambar 5.4734,17 33,59 2Gambar 5.4834,15 3Gambar 5.4933,58 Beras.jpg (Gambar 5.85) 1Gambar 5.4727,52 28,16 2Gambar 5.4827,07 3Gambar 5.4926,9 Batik.jpg (Gambar 5.87) 1Gambar 5.4722,33 22,42 2Gambar 5.4821,98 3Gambar 5.4921,89 Jeans.jpg (Gambar 5.89) 1Gambar 5.4717,46 16,49 2Gambar 5.4817,53 3Gambar 5.4917,2
30
Conclusion 0 The score difference between the program and Photoshop are not too far. It means, the output of the program is similar to Photoshop has. 0 The program does not process the “small particle” image too well because there are detail from the image (which is very small) that lost. (Ex: Kopi.jpg) 0 The program consumes high resources and takes a long time. The program runs about 3 minutes for each repetition 0 The repetition maximum amount is 3, otherwise the result image will be blurred.
31
Conclusion (cont.) 0 Image resolution also affects the result image. The result of low resolution image will be more blurred (compare to the high resolution image). 0 There are several types of images that are not suitable to use skeletonize process due to the residues that are not accurate and the effectiveness of time.
32
THANK YOU
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.