Region segmentation techniques for object-based image compression: a review

2004 ◽  
Author(s):  
Mark S. Schmalz ◽  
Gerhard X. Ritter
Keyword(s):  
Image Compression ◽  
Region Segmentation ◽  
Object Based

Error measures for object-based image compression

2005 ◽  
Author(s):  
Mark S. Schmalz ◽  
Gerhard X. Ritter
Keyword(s):  
Image Compression ◽  
Error Measures ◽  
Object Based

Object-based SAR image compression using vector quantization

2000 ◽  
Vol 36 (4) ◽  
pp. 1036-1046 ◽  
Author(s):  
N.M. Nasrabadi ◽  
H. Kwon ◽  
M. Venkatraman
Keyword(s):  
Image Compression ◽  
Vector Quantization ◽  
Sar Image ◽  
Object Based

scholarly journals Pengaruh Tingkat Kompresi Citra ALOS AVNIR-2 terhadap Akurasi Hasil Transformasi Indeks Vegetasi dan Klasifikasi Penutup Lahan Wilayah Salatiga dan Ambarawa, Jawa Tengah

2020 ◽  
Vol 34 (2) ◽  
pp. 130
Author(s):  
Projo Danoedoro
Keyword(s):  
Image Compression ◽  
Vegetation Index ◽  
Vegetation Indices ◽  
Original Data ◽  
Data Availability ◽  
Classification Methods ◽  
Quality Of Data ◽  
Pixel Classification ◽  
Central Java ◽  
Object Based

Abstrak Penggunaan teknik kompresi untuk menghemat ukuran penyimpanan citra digital telah banyak dijumpai dalam aplikasi keseharian. Di sisi lain, kompresi citra juga dapat memberikan konsekuensi berupa kehilangan detil data, yang akan berpengaruh pada integritas data. dan secara teoretis juga akan berpengaruh pada kualitas turunan data.  Penelitian ini mengkaji pengaruh tingkat kompresi citra digital multispektral ALOS-AVNIR2 yang terdiri dari empat saluran dengan resolusi spasial 10 meter terhadap akurasi hasil transformasi indeks vegetasi dan  klasifikasi penutup lahan untuk wilayah Salatiga-Ambarawa, Jawa Tengah.  Citra dikompresi pada sembilan tingkat, yaitu dari tidak kehilangan detil sama sekali (100%, atau sama dengan data asli) hingga 10%, dengan interval 10%. Indeks Vegetasi yang diterapkan meliputi NDVI, TVI dan MSARVI. Klasifikasi multispektral yang diujicobakan meliputi  klasifikasi per-piksel  dan klasifikasi berbasis objek.  Hasil penelitian ini menunjukkan bahwa transformasi indeks vegetasi dan klasifikasi per-piksel mengalami penurunan akurasi secara drastis, sejalan dengan meningkatnya kompresi citra, sementara klasifikasi berbasis objek mengalami perubahan akurasi relatif lebih sedikit dibandingkan analisis per-piksel. Temuan penelitian ini menunjukkan bahwa penggunaan citra terkompresi sebagai masukan proses klasifikasi secara digital sebaiknya dihindari. Meskipun demikian, kalau pun terpaksa dilakukan karena masalah ketersediaan data, maka metode klasifikasi berbasis objeklah yang sebaiknya diterapkan; dan untuk klasifikasi per-piksel maka algoritma jarak minimum terhadap rerata-lah yang  sebaiknya dipilih. Abstract The use of compression techniques for saving storage space of digital imagery has been commonly found in daily applications.  On the other hand, image compression can also provide consequences of losing data details, which will affect data integrity and theoretically will also affect the quality of data derived. This study examined the effect of ALOS-AVNIR2 multispectal image compression level consisting of four channels with 10 m spatial resolution to the accuracies of vegetation index transformation and land cover classification for Salatiga and Ambarawa region, Central Java. This study compressed the image into nine levels, i.e. from lossless details (100%, or equal to original data) up to 10% compression, at 10% intervals. The applied vegetation indices include NDVI, TVI and MSARVI. The multispectral classifications that were piloted include the per-pixel and object-based classification methods. The results of this study indicated that the vegetation index transformation and per-pixel classification have drastically decreased accuracies, in line with the increase in image compression; while the object-based classification has relatively more stable than per-pixel analysis. The findings of this study showed that the use of compressed imagery as an input to digital classification process should be avoided. However, even if it has to be done due to data availability issues, then object-based classification methods should be applied; and especially for per-pixel classification,  the minimum distance to mean algorithm should be chose.

Region Based Segmentation Techniques for Digital Images

2019 ◽  
Vol 16 (9) ◽  
pp. 3792-3801
Author(s):  
Wadood Abdul
Keyword(s):  
Image Segmentation ◽  
Image Compression ◽  
Frequency Response ◽  
Digital Images ◽  
Human Life ◽  
Region Segmentation ◽  
Image Prior ◽  
Scientific Disciplines ◽  
Entire Picture

This paper discusses region based segmentation techniques for digital images. For a few applications, such as image compression or recognition, we cannot handle the entire picture straightforwardly as it is unconventional and inefficient. Due to these reasons, many algorithms related to image segmentation are proposed in the literature to divide an image prior to compression or recognition. The segmentation of an image is basically done to arrange or group the image in a few fragments (districts) as specified by the elements of an image, for instance, according to the value of pixel or frequency response. Currently, many image segmentation approaches exist and are widely used in across scientific disciplines and daily human life. The segmentation approaches can be generally categorized to segmentation based on region, segmentation based on edges, and information grouping.

Sign in / Sign up

Export Citation Format

Share Document