@Z22 - the texture2D function performs interpolation, not simple sampling of the nearest full pixel.
In your N_Hance shader you have
vec2 Offset = vec2(1.5) / textureSize0; //Change according to resolution of girl. 1.5 for 3k , 2.5 for 1080p 3.5 for 720p, and 5.0 for 480p girls.
An observation I made was that the ratios and products of your factors and the vertical resolutions are
Vres . ,F. F/ Vres F*Vres
1440 1.5 0.0010 2160
1080 2.5 0.0023 2700
0720 3.5 0.0048 2520
0480 5.0 0.0104 2400
i.e. F*Vres is very approximately constant. As the values you used for F are rounded to the nearest 0.5 this suggests that this might be significant and that using more significant digits for F may make them more consistant. Arbitrarily assuming that the 1080p value for F is optimal (which I doubt is true) and normalising the others accordingly in a spreadsheet I get
Vres . ,F. ........ F/Vres........ F*Vres
1440 1.875 0.0013020833 2700
1080 2.500 0.0023148148 2700
0720 3.750 0.0052083333 2700
0480 5.625 0.0117187500 2700
or if I normalise to the 1440 vertical resolution I get
Vres . ,F. ........ F/Vres........ F*Vres
1440 1.500 0.0010416667 2160
1080 2.000 0.0018518519 2160
0720 3.000 0.0041666667 2160
0480 4.500 0.0093750000 2160
quite how to use this observation is another matter.