[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Re: Acquiring Dental RVG on Linux

On Wed, Jan 06, 2021 at 05:13:58PM +0530, Sonali Warunjikar wrote:
> I think distance calibration is over. Now over to other image parameters.

Trying to figure out appropriate post-processing that is needed / that the
proprietary application does to get the images right. Haven't got any
clues as yet. Following are some observations:

- For a recommended level of exposure the vendor application produces
  clearer images. For the same exposure our driver produces whitish
  images. Increasing the exposure gives better images on our system but
  that is not something we'd like to do.

- As Karsten advised, an attempt was made to get usbmon trace of Windows X
  ray and process the same on both systems. But there is a repeatable and
  mysterious loss of 1 chunk of 57856 bytes when collecting data this way.
  Not only the missing bytes but their location is also unclear keeping me
  from reproducing the same image. Nevertheless we can take x ray of the
  same area on both systems as the difference is to be judged mostly

- Our driver gets the data in 2 Ci interactions whereas the vendor app
  makes several interactions to get data in 3 different chunk sizes in 111
  36672*1 + 57856*54 + 7680*56 = 3590976
  (Total bytes expected are 3648832 and 1 chunk of 57856 is missing in
  usbmon as mentioned above).

  Not sure why 111 interactions and why these chunk sizes when 2
  interactions are sufficing. But hope it has nothing to do with image

  In both systems there is no perceivable difference in the amount of time
  for these interactions.

  In both systems the returning of data starts with a delay of 2 to 3
  seconds after exposure. So if the sensor needs some time for its data to
  settle, it's taking the same on both systems before starting returning

  Overall I am inclined to believe that there is no problem in the device
  interaction. We have to just figure out suitable post processing.

- Simple normalizations to utilize the full contrast and reduction in
  brightness or any other manual adjustment to contrast and brightness
  does not make up for less exposure in our application, whereas vendor
  application manages to get it right with the same exposure level.

- Gimp has an exposure adjustment filter, which does not help in this
  case. It helps only as a refinement if the image is already somewhat

- It does not look like any non linear adjustment is required - at least
  at this stage when something very coarse seems to be missing.

Any pointers on suitable filters for such post processing would be very

Reply to: