|
|
Processing of high-resolution retinal images
Vraňáková, Sofia ; Chmelík, Jiří (oponent) ; Valterová, Eva (vedoucí práce)
The master’s thesis is focused on the processing of high-resolution retinal images. The aim of this work is to obtain images of higher quality from a sequence of low-quality frames. The frames are first pre-processed by using bilateral filtering and contrast enhancement. The shift between the frames in the imaging sequence is estimated using phase correlation, and these frames are then fused together using the averaging through the frames and the super-resolution technique, more specifically regularization based on bilateral total variance. The resulting median quality scores of obtained images are PIQUE 0.2600, NIQE 0.0701, and BRISQUE 0.3936 for the averaging technique and PIQUE 0.1063, NIQE 0.0507, and BRISQUE 0.1570 for super-resolution technique.
|
|
|
Pedometer and fall detection
Vraňáková, Sofia ; Vítek, Martin (oponent) ; Němcová, Andrea (vedoucí práce)
This work is focused on the issue of human activity sensing using inertial sensors built into intelligent devices and followed by detection of walk, detection of steps and fall from the accelerometer-scanned signal in the smartphone. The walk detection is performed by using signal envelope, a first signal difference and a standard deviation of the signal. The step detection is implemented by using the method of searching local maximum values and the customized filtering method. Fall detection is realized by using the method of searching maximum and minimum values and using the appropriate threshold values.
|
|
|
Processing of high-resolution retinal images
Vraňáková, Sofia ; Chmelík, Jiří (oponent) ; Valterová, Eva (vedoucí práce)
The master’s thesis is focused on the processing of high-resolution retinal images. The aim of this work is to obtain images of higher quality from a sequence of low-quality frames. The frames are first pre-processed by using bilateral filtering and contrast enhancement. The shift between the frames in the imaging sequence is estimated using phase correlation, and these frames are then fused together using the averaging through the frames and the super-resolution technique, more specifically regularization based on bilateral total variance. The resulting median quality scores of obtained images are PIQUE 0.2600, NIQE 0.0701, and BRISQUE 0.3936 for the averaging technique and PIQUE 0.1063, NIQE 0.0507, and BRISQUE 0.1570 for super-resolution technique.
|
|
|
Pedometer and fall detection
Vraňáková, Sofia ; Vítek, Martin (oponent) ; Němcová, Andrea (vedoucí práce)
This work is focused on the issue of human activity sensing using inertial sensors built into intelligent devices and followed by detection of walk, detection of steps and fall from the accelerometer-scanned signal in the smartphone. The walk detection is performed by using signal envelope, a first signal difference and a standard deviation of the signal. The step detection is implemented by using the method of searching local maximum values and the customized filtering method. Fall detection is realized by using the method of searching maximum and minimum values and using the appropriate threshold values.
|
host ::
RSS.