I have just set off a render of the entire map that will be 10000 x 5422pi, using Monsen's idea of giving the coordinates rather than trying to hit the spot with the cursor when defining the print rectangle. This is the second attempt to see how far I can push the limits in a single image. I forgot the anti-aliasing the first time.
When I did this the first time, however the render took only 17 passes. This time, with identical coordinates, its taking 21 passes. Is that just because I ticked the anti-aliasing box and gave it 10 percent, does anyone know?
The anti-alias will do that, yes. For the antialias to work, it needs to export a larger map than the final result, so basically, with a 10% antialias, the true export will be 10% extra in each dimension compared to what you requested. This also means that the maximum image size you will be able to export will be less with antialias enabled.
It worked, and now I can safely say that I will need to make a tiled image that (if it were stuck together in one big map) would be at least 30000 pixels in width to capture the smallest details. To keep the tiles reasonably sized for quick loading etc, I would need to create tiles that are approximately 1000 pixels on the longest side, which means I would need to cut each side into 30 tiles, which means... 900 tiles?
I think I will do the tiling just once, right at the end... and I can see why WarEagle developed an automated procedure for sorting this out with far less bother
This is about 1/4 of one of my 'draft' map tiles - a test for the final effort (I have decided to cheat a bit and only do a 15 x 15 tile grid, which is 225 maps, instead of 900. The up side of doing tiles is that they only take a few seconds to render each time).
Now that everything is pin sharp there are clear differences between the resolutions of the bitmap fills, and the resolutions of the vector objects - the rivers and lakes.
I think, that since it would be near impossible to 'sharpen the textures', I should deliberately blur the vector objects just a tiny, tiny fraction, so that its hardly noticeable - just so they don't jump out at you at large magnifications. The alternative is to remove the bitmap fills and do the job entirely with symbols, which would just about kill my laptop I think, if not me!
I understand the thing about having obviously different resolutions of source material. Instead of blurring the sharp elements to reduce their effective resolution, another option would be to actually reduce the resolution of everything. Output at half or a third of the above resolution and scale everything up again using a paint-type program. The advantage of what I describe is that everything will be rendered to roughly the same resolution and will have roughly the same amount of blur to it (which may or may not be desirable).
I have found that I get fairly good results at a two-times upscale in most cases, but things start to degrade a bit from there. Above about a three-times resize, I find that a wavelet or fractal image resizer (there are a whole lot of them running around online) can often give good enough results to get by (they may also be labeled as a "smart resizer"). This soft of processing doesn't actually put back the original information lost when you downsampled everything, but it does tell enough lies of the right sort that your visual system doesn't care too much about it most of the time.
At the risk of boring everyone... I've been working on bringing the resolution of different source materials together a little better, which has included reducing the scale of the forest texture, increasing the scale of the water texture, adding an outer glow to the rivers and lakes to make them blend a little better with the land, redrawing the forest to give it more edge detail, pulling it back from hanging too far out over the edge of the cliff (its on top of the symbol sheet in this instance), and learning to accept that if I want to use the mountain symbols the size I like them (which is large), they are always going to look more pixelated than anything else. I'm hoping this will just become part of the nature of the map, rather than a distraction.
Its not that, Jim, I just can't work out the tiles properly. lol. I'm no mathematician, so I work things like this out by visualising them in my head, but my head isn't big enough to hold onto the entire problem all at the same time, when the width of the map is exactly 1.844011 times the height, rather than a nice round number like 2. I keep confusing myself when converting between map units and rectangular section PNG output dimensions (which are in pixels). This means I end up with unintended over and underlaps all over the place. Things would be simpler if the map was equi-rectangular, rather than Miller Cylindrical in projection, because then at least I would be able to base everything on there being two perfect square areas side by side and work from there. The fact is, however, that Errispa 3 is a Miller, and there isn't a way to convert what I've already done to make it become an equi-rectangular.
Put simply. I have to start all over again with the right kind of projection to make it possible for me to get my head around the conversions between map units and pixels. The landmass, I'm afraid, doesn't look half as good when its projected as an equi-rectangular, so I'm also going to have to pick another world before I get started.
I may be some time, but at least you know I haven't completely given up
Shame really. I did a really excellent spreadsheet that automatically calculated and concatenated all the exact coordinates in map units for all 225 tiles. I suppose it will come in useful again someday. You only have to input the width and height of the map in map units and it does the rest for you.
I don't know, Dogtag. I hadn't thought of doing it that way, but I need a really high total resolution - something in the range of 30,000 pixels in width, just to make the detail clear at maximum zoom level. I'll go and have a look see now, shall I?
No. I'm afraid it doesn't, but its always worth checking an idea.
EDIT: Well - it does, but only to pdf - and that's only because Win 10 comes with a built in print to pdf facility... which automatically reduces the resolution. lol. Win 10 is such a 'nanny' about everything. I feel like I've been bound and gagged hand and foot. Bring back Windows XP, or earlier!
If anyone with a better mathematical brain wants to have a go, and also happens to have a recent version of Excel on their PC, this spreadsheet automatically calculates the coordinates for a grid of tiles 15 x15, even if the map isn't square. All you have to do is input the width and height of the map in map units at the end of the first table. I increased the dimensions of Errispa 3 (still shown in the table) by a couple of map units (no more than 10 in over 20,000), to simplify matters - to ensure the rest of the table was integers only. (It gets really confusing if you have to input coordinates with multiple decimal places). All this means is that there would have been a strip of white down the far end and along the top of the map, but on a white background this really wouldn't have mattered.
This is Errispa 4. I know the coastline is vaguely reminiscent of a running rabbit, I think this illusion will fade once I start adding things to the land. Everyone already knows I need a swamp, a cliff and a waterfall, lol, but the white lines are a different technique I've decided to use to dictate the presence and position of mountain chains. I look for lines of land that protrude from the coast, as if there is more bulk to the land in those places, and sketch the white lines to pick them out. These, I hope, will prove to be more realistic ranges than before.
There is also a grid - 36 x 18 - each grid square representing 10 degrees of latitude and longitude, to give both myself and anyone looking at the map a greater sense of size and relative position. I'm hoping this will also help when I get to the point where I'm setting about creating the tiles I will need to make the zomable map.
Comments
When I did this the first time, however the render took only 17 passes. This time, with identical coordinates, its taking 21 passes. Is that just because I ticked the anti-aliasing box and gave it 10 percent, does anyone know?
I'd better add this comment before it goes pop!
I think I will do the tiling just once, right at the end... and I can see why WarEagle developed an automated procedure for sorting this out with far less bother
Now that everything is pin sharp there are clear differences between the resolutions of the bitmap fills, and the resolutions of the vector objects - the rivers and lakes.
I think, that since it would be near impossible to 'sharpen the textures', I should deliberately blur the vector objects just a tiny, tiny fraction, so that its hardly noticeable - just so they don't jump out at you at large magnifications. The alternative is to remove the bitmap fills and do the job entirely with symbols, which would just about kill my laptop I think, if not me!
What do you think?
I have found that I get fairly good results at a two-times upscale in most cases, but things start to degrade a bit from there. Above about a three-times resize, I find that a wavelet or fractal image resizer (there are a whole lot of them running around online) can often give good enough results to get by (they may also be labeled as a "smart resizer"). This soft of processing doesn't actually put back the original information lost when you downsampled everything, but it does tell enough lies of the right sort that your visual system doesn't care too much about it most of the time.
Sorry!
Put simply. I have to start all over again with the right kind of projection to make it possible for me to get my head around the conversions between map units and pixels. The landmass, I'm afraid, doesn't look half as good when its projected as an equi-rectangular, so I'm also going to have to pick another world before I get started.
I may be some time, but at least you know I haven't completely given up
EDIT: Well - it does, but only to pdf - and that's only because Win 10 comes with a built in print to pdf facility... which automatically reduces the resolution. lol. Win 10 is such a 'nanny' about everything. I feel like I've been bound and gagged hand and foot. Bring back Windows XP, or earlier!
This is Errispa 4. I know the coastline is vaguely reminiscent of a running rabbit, I think this illusion will fade once I start adding things to the land. Everyone already knows I need a swamp, a cliff and a waterfall, lol, but the white lines are a different technique I've decided to use to dictate the presence and position of mountain chains. I look for lines of land that protrude from the coast, as if there is more bulk to the land in those places, and sketch the white lines to pick them out. These, I hope, will prove to be more realistic ranges than before.
There is also a grid - 36 x 18 - each grid square representing 10 degrees of latitude and longitude, to give both myself and anyone looking at the map a greater sense of size and relative position. I'm hoping this will also help when I get to the point where I'm setting about creating the tiles I will need to make the zomable map.