Mask Operations Crash?

[dvhirst]

Hello Jonathan,

I am working with a pano image that is 21000x3000 pel. I had good success with the Mask>Separate on 2 colors transform, and got a good place to start building a mask that I want to use with the Haze filter. When I try to fill in the lower part of the image, I was fairly successful using Mask>Paint (fixed similar pixels). The app crashed, but I was able to recover most of my work. Then I tried using Mask>FloodFill, that also worked, until it didn't. The app crashed repeatedly and recovery of work to point of crash has become a lengthy process. This image is after recovering the latest crash, which did recover successfully to the point of the crash. Now, I am filling in the bottom section of the mask, in preparation for using the Haze filter.

Image 1418 (1024x768).jpg (95.61 KiB) Viewed 1871 times

My questions:

1. Is my image too large; perhaps I should segment my image and work on it a segment at a time?
2. Will saving the current mask help the process go more smoothly?
3. Do you have any suggestions for a better approach to building the 2-part black/white mask I need?

Thanks.

[jsachs]

It sounds like you are running out of memory. PWP keeps many copies of your images around in memory and a 21000x3000 48-bit color image uses about 380MB of RAM per copy and 63MB per mask. Each mask operation stores a couple of copies of the mask so as you build a complicated mask it adds up fast.

How much RAM do you have on your computer? To work with images of that size, I would recommend at least 16 and preferably 32 GB. Even if the images will fit, there is a lot of overhead for Windows and all the other apps running in the background.

[dvhirst]

I suspected as much. I have 16 GB of main memory and 32 GB of virtual disk memory on a dedicated partition. Occupancy, when the app crashes, is at 97% and it seems to crash when I attempt an undo operation. So, if I commit the edits every few operations, will that reduce the memory requirement or would I be better off to segment the image and work on a segment at a time? Moving to 32 GB of main memory means a new MoBo, and that's not in the cards. Here's my current config:

RYO System Configuration (updated 2020 08): SCP: Win10.1909.18363.959 Pro 64bit; Mobo: ASUS Crosshair IV Formula Rev 1.02G; BIOS: American Megatrends Inc. V1005 08/06/2010; CPU: AMD Phenom II X4 965 C3 BE, 3.4 GHz; RAM: Corsair Dominator DDR3, 16 GB (4 x 4 GB); SATA III SSD: Samsung 860 EVO 500GB; SATA III HDD: 3 x Seagate 2 TB (Hybrid SSD/HDD), 1 x WD 500 GB, 3 x Seagate Backup Plus 3TB USB3 (with Seagate Replica 5 SW for automatic backup); ODD: 1 x Plextor SATA DVD RW, 1 x ASUS SATA BlueRay DVD RW; VGA: EVGA GEForce GTX 750, 1 GB; USB3: 7 port PCIe card, PSU: OCZ FirePower Tech 550W; Case: Eagle Tech-Skyhawk Server; Cooling: CPU-stock; 2 case fans.

[jsachs]

I would recommend doing one or two operations at a time, saving the image, clearing the workspace, reloading the image and continuing. This will limit the number of copies of the image that have to be in memory at one time. Unfortunately, it makes it much harder to change something at the beginning of the process once you have moved on to the next step.

You might also consider resizing the image down by a factor of two which will reduce the memory requirement by a factor of 4. Most images are not really that sharp when viewed at 100%, so you may be able to make as large a print from an image with fewer, but higher quality pixels as from the original. Also, once you make your main tonal adjustments, converting to 24-bit color will make little difference and will halve the size of the images.

[dvhirst]

I've been thinking about my process here. The 2-Color Separation step went through without any issues and I got a quite useable mask from that step. Now, I am working on the mask itself as a .tif image. I've just reloaded the mask, and with nothing else going on in my server, PWP has a memory occupancy of 39 MB. My next step will be to reload the mask image from the time of the crash. I'll take a look a memory usage then, and work through your steps, keeping an eye on memory occupancy. My base image is a jpeg image. I could go all the way back to my raw files and start over with a .tif image as the starting point, I was trying to avoid that if I can. The jpeg image should be good enough. Thanks.