Action-Packed Weekend at Lake Coeur d’Alene

Adventure Log
5

So much awesome stuff here - action packed indeed!

Fun to live vicariously through your adventures, especially since I’m stuck inside at the moment.

I really enjoyed the umbrella rescue, and seeing the fish darting in an out at various points. Was also cool to see the thermocline layer effects - would be fun to see the temperature vs depth graph there :slight_smile:
So much content makes me wonder how hard it would be to automate some kind of montage creation, or auto fast-forwarding between the main events :smiley:

I feel like every time I watch an ROV video it gets me thinking about the processing required to improve the visibility. Normalising each colour channel and adding some extra contrast can help with seeing more detail (at the cost of some flickering when brightness levels change):

#!/usr/bin/env python3
import cv2
import numpy as np

def process(frame):
    result = np.empty_like(frame)
    # normalise each colour channel
    for channel in range(3):
        result[:,:,channel] = cv2.normalize(frame[:,:,channel], None, 0, 255, cv2.NORM_MINMAX)
    # clip off the bottom 10% of values to increase contrast
    return cv2.convertScaleAbs(result, None, 1.1, -25)

processed
processed2664×1000 186 KB

processed2
processed21380×1206 224 KB

but there are definitely more advanced but still quite lightweight algorithms that could do a better job. Then there’s actual attempts at colour correction, which is perhaps also valuable/important for different purposes, and would require a different approach.

Oh also, have you tried the opacity option in the recent PingViewer update? Could be helpful to see some extra video in that top right corner :slight_smile:

2 Likes